WEBVTT 00:00:00.000 --> 00:00:02.460 align:middle line:90% [SQUEAKING] 00:00:02.460 --> 00:00:04.428 align:middle line:90% [RUSTLING] 00:00:04.428 --> 00:00:07.380 align:middle line:90% [CLICKING] 00:00:07.380 --> 00:00:16.730 align:middle line:90% 00:00:16.730 --> 00:00:18.300 align:middle line:84% PROFESSOR: OK, so let me get started. 00:00:18.300 --> 00:00:21.380 align:middle line:84% So if you remember last class, I did 00:00:21.380 --> 00:00:24.200 align:middle line:84% the basic strategic investment theory 00:00:24.200 --> 00:00:29.780 align:middle line:84% and talked about how we think about firms trying 00:00:29.780 --> 00:00:34.580 align:middle line:84% to make long term investments that either deter rivals 00:00:34.580 --> 00:00:38.403 align:middle line:84% from entering or changing the nature of competition 00:00:38.403 --> 00:00:40.820 align:middle line:84% in a way that's going to make them more profitable if they 00:00:40.820 --> 00:00:42.920 align:middle line:84% go ahead and give up and accommodate that the rivals come 00:00:42.920 --> 00:00:43.560 align:middle line:90% in. 00:00:43.560 --> 00:00:47.450 align:middle line:84% And the basic insights there are, to deter entry, 00:00:47.450 --> 00:00:48.270 align:middle line:90% it's fairly simple. 00:00:48.270 --> 00:00:49.760 align:middle line:84% You just want to do whatever kinds of investments 00:00:49.760 --> 00:00:51.300 align:middle line:90% are going to hurt your rivals. 00:00:51.300 --> 00:00:54.260 align:middle line:84% And then when you're talking about accommodating entry, 00:00:54.260 --> 00:00:56.258 align:middle line:84% one of the insights was it really depends on 00:00:56.258 --> 00:00:58.550 align:middle line:84% whether you're playing games with strategic complements 00:00:58.550 --> 00:00:59.520 align:middle line:90% or substitutes. 00:00:59.520 --> 00:01:01.400 align:middle line:84% And in strategic complements games, 00:01:01.400 --> 00:01:05.060 align:middle line:84% generally investments that help your rivals 00:01:05.060 --> 00:01:06.830 align:middle line:84% are ones that are good strategically, 00:01:06.830 --> 00:01:09.300 align:middle line:84% and investments that hurt your rivals are bad. 00:01:09.300 --> 00:01:12.980 align:middle line:84% And then it's reversed if you're looking at Cournot wide games. 00:01:12.980 --> 00:01:16.980 align:middle line:84% So today, I'm going to talk about empirical work 00:01:16.980 --> 00:01:18.430 align:middle line:90% on strategic investment. 00:01:18.430 --> 00:01:21.120 align:middle line:84% And why are we interested in this? 00:01:21.120 --> 00:01:24.060 align:middle line:84% There's the standard considerations 00:01:24.060 --> 00:01:25.390 align:middle line:90% that we want to understand. 00:01:25.390 --> 00:01:26.470 align:middle line:90% Do the models work well? 00:01:26.470 --> 00:01:28.110 align:middle line:90% Do firms do what we think? 00:01:28.110 --> 00:01:32.340 align:middle line:84% There are other cases where strategic investment 00:01:32.340 --> 00:01:33.990 align:middle line:84% of the type described in those models 00:01:33.990 --> 00:01:35.790 align:middle line:84% could be violating antitrust laws. 00:01:35.790 --> 00:01:38.940 align:middle line:84% I gave a model of tying where tying 00:01:38.940 --> 00:01:40.650 align:middle line:90% could be used to deter entry. 00:01:40.650 --> 00:01:43.616 align:middle line:84% That would potentially be a monopolization violation. 00:01:43.616 --> 00:01:48.640 align:middle line:90% 00:01:48.640 --> 00:01:52.620 align:middle line:84% So anyway, so yeah, trying to see, do these-- 00:01:52.620 --> 00:01:54.630 align:middle line:84% and are the incentives that these models 00:01:54.630 --> 00:01:57.132 align:middle line:84% say could be there to signal costs or do other things? 00:01:57.132 --> 00:01:58.840 align:middle line:84% Are those actually empirically important? 00:01:58.840 --> 00:02:02.040 align:middle line:84% And then if they are important, do the firms recognize and react 00:02:02.040 --> 00:02:03.070 align:middle line:90% to them? 00:02:03.070 --> 00:02:07.270 align:middle line:84% So if we're thinking about, one, are these incentives important, 00:02:07.270 --> 00:02:10.170 align:middle line:84% the main thing to understand is this d a1 star 00:02:10.170 --> 00:02:14.010 align:middle line:90% / dk1 and da2 star / dk1. 00:02:14.010 --> 00:02:17.520 align:middle line:84% If you remember, when we're talking about the entry 00:02:17.520 --> 00:02:22.790 align:middle line:84% deterrence, whether you do more or less investment 00:02:22.790 --> 00:02:25.370 align:middle line:84% to deter entry was determined by a strategic effect. 00:02:25.370 --> 00:02:27.515 align:middle line:84% And that strategic effect looked like-- 00:02:27.515 --> 00:02:30.500 align:middle line:90% 00:02:30.500 --> 00:02:32.673 align:middle line:90% it was d pi 1-- 00:02:32.673 --> 00:02:36.540 align:middle line:90% 00:02:36.540 --> 00:02:42.945 align:middle line:84% sorry, d pi 2 / da1, da1 star / dk1. 00:02:42.945 --> 00:02:46.050 align:middle line:90% 00:02:46.050 --> 00:02:47.730 align:middle line:84% Strategic entry deterrence was about 00:02:47.730 --> 00:02:53.590 align:middle line:84% k1 is a commitment that makes me change my second period action. 00:02:53.590 --> 00:02:55.800 align:middle line:84% And then when I change my second period action, 00:02:55.800 --> 00:03:00.540 align:middle line:84% how does it affect firm 2 and keep firm 2 out of the market? 00:03:00.540 --> 00:03:05.100 align:middle line:84% Accommodation, I went to the helping and hurting. 00:03:05.100 --> 00:03:07.290 align:middle line:84% But before that, I had the strategic effect 00:03:07.290 --> 00:03:14.460 align:middle line:84% was d pi 1 / da2, da2 star / dk1. 00:03:14.460 --> 00:03:17.310 align:middle line:84% So strategic entry accommodation was 00:03:17.310 --> 00:03:21.000 align:middle line:84% about getting your rival to do something that-- using k1 00:03:21.000 --> 00:03:23.200 align:middle line:84% as a change in the nature of competition 00:03:23.200 --> 00:03:25.033 align:middle line:84% in a way that get your rival to do something 00:03:25.033 --> 00:03:26.400 align:middle line:90% that you want to do. 00:03:26.400 --> 00:03:28.830 align:middle line:84% We pretty much always know what these are-- the signs 00:03:28.830 --> 00:03:30.430 align:middle line:90% of these, d pi 2 / da1. 00:03:30.430 --> 00:03:32.940 align:middle line:84% This is just like firms like their opponents 00:03:32.940 --> 00:03:35.610 align:middle line:84% to set higher prices, or firms like their opponents 00:03:35.610 --> 00:03:37.230 align:middle line:90% choose lower quantities. 00:03:37.230 --> 00:03:38.980 align:middle line:84% So then, the important strategic questions 00:03:38.980 --> 00:03:40.780 align:middle line:84% are, how big are these things that-- 00:03:40.780 --> 00:03:45.520 align:middle line:84% how much do the a's change with the investments? 00:03:45.520 --> 00:03:48.543 align:middle line:84% So anyway, that's one thing people do. 00:03:48.543 --> 00:03:49.960 align:middle line:84% And then the other thing people do 00:03:49.960 --> 00:03:57.060 align:middle line:84% is we're trying to see, does the model with these investment-- 00:03:57.060 --> 00:03:59.290 align:middle line:84% entry deterring investments better explain 00:03:59.290 --> 00:04:01.800 align:middle line:84% what firms do than a model without those investment 00:04:01.800 --> 00:04:07.230 align:middle line:84% firms in there, and are firms recognizing those? 00:04:07.230 --> 00:04:10.470 align:middle line:84% Sometimes, we can think of this question 00:04:10.470 --> 00:04:13.470 align:middle line:84% of are firms recognizing and reacting to entry deterrence 00:04:13.470 --> 00:04:15.510 align:middle line:84% incentives by really thinking of it as just 00:04:15.510 --> 00:04:17.579 align:middle line:90% different strategic models. 00:04:17.579 --> 00:04:19.829 align:middle line:84% So as I said, everything I did on strategic investment 00:04:19.829 --> 00:04:23.160 align:middle line:84% was comparing the investment with an entry accommodation 00:04:23.160 --> 00:04:25.990 align:middle line:84% motive to the open loop equilibrium. 00:04:25.990 --> 00:04:29.190 align:middle line:84% And the open loop equilibrium is the fully optimal choice 00:04:29.190 --> 00:04:33.480 align:middle line:84% of k in this model, where K is never observed. 00:04:33.480 --> 00:04:37.690 align:middle line:84% The incumbent chooses Kk The entrant chooses in or out. 00:04:37.690 --> 00:04:41.280 align:middle line:84% There's monopoly or duopoly, and only after the-- 00:04:41.280 --> 00:04:45.120 align:middle line:84% and never does the rival observe k. 00:04:45.120 --> 00:04:47.610 align:middle line:84% So you can think of non-strategic investment 00:04:47.610 --> 00:04:50.370 align:middle line:84% as just strategic investment in that model 00:04:50.370 --> 00:04:51.820 align:middle line:90% where k is unobserved. 00:04:51.820 --> 00:04:55.220 align:middle line:84% And then you could add in two different strategic variants, 00:04:55.220 --> 00:04:59.830 align:middle line:84% one where if you observed k here, 00:04:59.830 --> 00:05:01.690 align:middle line:84% that's the fully strategic model where 00:05:01.690 --> 00:05:03.640 align:middle line:84% you think about entry accommodation and entry 00:05:03.640 --> 00:05:04.880 align:middle line:90% deterrence motives. 00:05:04.880 --> 00:05:07.120 align:middle line:84% And if you observe k here, that's 00:05:07.120 --> 00:05:09.850 align:middle line:84% the model where you only think about strategic entry 00:05:09.850 --> 00:05:12.430 align:middle line:84% accommodation, not deterrence, because you're not observing k 00:05:12.430 --> 00:05:13.845 align:middle line:90% until after the entry decision. 00:05:13.845 --> 00:05:16.950 align:middle line:90% 00:05:16.950 --> 00:05:19.080 align:middle line:84% Anyway, so I'm going to talk about papers taking 00:05:19.080 --> 00:05:20.750 align:middle line:90% a few different approaches. 00:05:20.750 --> 00:05:24.480 align:middle line:84% First approach is just examine the effects of the investment 00:05:24.480 --> 00:05:25.120 align:middle line:90% directly. 00:05:25.120 --> 00:05:27.600 align:middle line:90% Just look at some investment, k. 00:05:27.600 --> 00:05:29.880 align:middle line:84% Look at some action, a, that occurs later, 00:05:29.880 --> 00:05:34.050 align:middle line:84% and observe whether da / dk is large, and what the sign of it 00:05:34.050 --> 00:05:35.070 align:middle line:90% is. 00:05:35.070 --> 00:05:39.570 align:middle line:84% Second approach you see is to try to understand if firms are 00:05:39.570 --> 00:05:41.940 align:middle line:84% following their strategic incentives, what we could do 00:05:41.940 --> 00:05:44.460 align:middle line:84% is examine reduced form predictions that 00:05:44.460 --> 00:05:46.860 align:middle line:84% differ between the non-strategic and strategic models, 00:05:46.860 --> 00:05:49.680 align:middle line:84% and then see which one better explains the data. 00:05:49.680 --> 00:05:53.280 align:middle line:84% And then the third approach is to write down a model 00:05:53.280 --> 00:05:55.977 align:middle line:84% that you can structurally estimate and understand, 00:05:55.977 --> 00:05:58.560 align:middle line:84% does a structural model explain things better than, hopefully, 00:05:58.560 --> 00:06:01.890 align:middle line:84% some alternate structural model that doesn't have that term? 00:06:01.890 --> 00:06:04.380 align:middle line:84% And so I'll talk about papers with each approach. 00:06:04.380 --> 00:06:08.460 align:middle line:84% So I'm going to start with Chevalier's paper, Capital 00:06:08.460 --> 00:06:10.950 align:middle line:84% Structure and Product Market Competition. 00:06:10.950 --> 00:06:13.440 align:middle line:84% So this is a paper about the supermarket industry, 00:06:13.440 --> 00:06:19.200 align:middle line:84% and it's motivated by work in corporate finance. 00:06:19.200 --> 00:06:21.870 align:middle line:84% So the corporate finance question 00:06:21.870 --> 00:06:24.840 align:middle line:84% that motivates this is, why do firms choose the capital 00:06:24.840 --> 00:06:26.700 align:middle line:90% structures that they do? 00:06:26.700 --> 00:06:27.540 align:middle line:90% I think we're used-- 00:06:27.540 --> 00:06:28.332 align:middle line:90% well, I don't know. 00:06:28.332 --> 00:06:30.790 align:middle line:84% How many people here have taken a corporate finance course? 00:06:30.790 --> 00:06:32.160 align:middle line:90% Is the answer going to be 0? 00:06:32.160 --> 00:06:34.620 align:middle line:90% OK, so-- one. 00:06:34.620 --> 00:06:35.220 align:middle line:90% Good. 00:06:35.220 --> 00:06:39.030 align:middle line:84% OK, anyway, so starting point for most theoretical 00:06:39.030 --> 00:06:43.330 align:middle line:84% corporate finance courses is you start with Miller- Modigliani. 00:06:43.330 --> 00:06:45.330 align:middle line:84% A lot of corporate finance is about how does a-- 00:06:45.330 --> 00:06:48.090 align:middle line:84% well, the basic structure that you certainly want to teach MBAs 00:06:48.090 --> 00:06:53.310 align:middle line:84% is, how do you set up your capital structure of your firm? 00:06:53.310 --> 00:06:55.800 align:middle line:84% How do you decide between raising money through debt, 00:06:55.800 --> 00:06:57.510 align:middle line:90% raising money through equity? 00:06:57.510 --> 00:07:01.830 align:middle line:84% How is a firm value affected by choosing a capital structure? 00:07:01.830 --> 00:07:04.020 align:middle line:84% And so the starting point for this literature 00:07:04.020 --> 00:07:06.610 align:middle line:84% is what's called the Miller-Modigliani theorem. 00:07:06.610 --> 00:07:08.070 align:middle line:84% And the Miller-Modigliani theorem 00:07:08.070 --> 00:07:09.840 align:middle line:84% is a basic theorem talking about how 00:07:09.840 --> 00:07:14.370 align:middle line:84% in competitive markets with different states of the world, 00:07:14.370 --> 00:07:17.008 align:middle line:84% your capital structure is just completely irrelevant. 00:07:17.008 --> 00:07:18.550 align:middle line:84% And if you think about it, it's like, 00:07:18.550 --> 00:07:21.010 align:middle line:84% there's just some giant set of states 00:07:21.010 --> 00:07:23.440 align:middle line:84% of the world that may arise, and the firm 00:07:23.440 --> 00:07:26.060 align:middle line:84% earns some different amount of profits in every state. 00:07:26.060 --> 00:07:29.320 align:middle line:84% And then you divide up the state contingent claims 00:07:29.320 --> 00:07:32.350 align:middle line:84% on the firms into a debt contract, 00:07:32.350 --> 00:07:35.480 align:middle line:84% or an equity contract-- like any contract, like equity, 00:07:35.480 --> 00:07:37.480 align:middle line:84% or debt, or whatever, pays off different amounts 00:07:37.480 --> 00:07:39.110 align:middle line:84% in different states of the world. 00:07:39.110 --> 00:07:41.770 align:middle line:84% And then Miller-Modigliani theorem is basically saying, 00:07:41.770 --> 00:07:43.520 align:middle line:84% look, if you think about it, if you're just deciding how much 00:07:43.520 --> 00:07:44.620 align:middle line:84% to have in different states of the world, 00:07:44.620 --> 00:07:46.995 align:middle line:84% whether you have some this combination of debt and equity 00:07:46.995 --> 00:07:50.205 align:middle line:84% or this combination, all you're doing is changing the debt 00:07:50.205 --> 00:07:52.330 align:middle line:84% contract or something, the equities were something, 00:07:52.330 --> 00:07:54.700 align:middle line:90% but the sum of the two is-- 00:07:54.700 --> 00:07:57.070 align:middle line:84% or the sum of all the claims you issue on the firm 00:07:57.070 --> 00:07:59.643 align:middle line:84% are still just issued to all of the value of the state 00:07:59.643 --> 00:08:01.310 align:middle line:84% contingent profits you're going to earn. 00:08:01.310 --> 00:08:04.990 align:middle line:84% So capital structure is just completely irrelevant. 00:08:04.990 --> 00:08:07.180 align:middle line:84% So then obviously, after that first observation, 00:08:07.180 --> 00:08:10.570 align:middle line:84% you need theories for why firms do 00:08:10.570 --> 00:08:13.370 align:middle line:84% have debt, and do have equity, and why they decide to do it. 00:08:13.370 --> 00:08:16.060 align:middle line:84% And I wouldn't say this is the biggest class, 00:08:16.060 --> 00:08:18.640 align:middle line:84% but a substantial part of that literature does say, 00:08:18.640 --> 00:08:21.610 align:middle line:84% one reason why you could want capital structure could matter 00:08:21.610 --> 00:08:25.060 align:middle line:84% is because capital structure can affect the product market 00:08:25.060 --> 00:08:26.780 align:middle line:90% competition between the firms. 00:08:26.780 --> 00:08:29.200 align:middle line:84% And there's a big literature about debt, 00:08:29.200 --> 00:08:33.460 align:middle line:84% arguing how debt could affect real actions that firms take, 00:08:33.460 --> 00:08:36.490 align:middle line:84% and effects that can go in different directions. 00:08:36.490 --> 00:08:40.059 align:middle line:84% So let's think about, thought experiment, a firm takes 00:08:40.059 --> 00:08:41.200 align:middle line:90% on a large amount of debt. 00:08:41.200 --> 00:08:43.600 align:middle line:84% And so because they've taken on a large amount of debt, 00:08:43.600 --> 00:08:46.210 align:middle line:84% they need a lot of cash flow to make the debt payments 00:08:46.210 --> 00:08:48.610 align:middle line:90% that they've taken on. 00:08:48.610 --> 00:08:49.940 align:middle line:90% What does that do? 00:08:49.940 --> 00:08:53.860 align:middle line:84% Well, first, start with a standard oligopoly game. 00:08:53.860 --> 00:08:55.610 align:middle line:84% That's going to have no effect whatsoever. 00:08:55.610 --> 00:08:57.500 align:middle line:90% So I take it on debt. 00:08:57.500 --> 00:08:58.850 align:middle line:90% I want to make more money. 00:08:58.850 --> 00:09:00.820 align:middle line:84% If it's a static game, every period 00:09:00.820 --> 00:09:02.590 align:middle line:84% I'm trying to make as much money as I can. 00:09:02.590 --> 00:09:04.882 align:middle line:84% So I'm still just going to choose the profit maximizing 00:09:04.882 --> 00:09:06.370 align:middle line:90% prices in every period. 00:09:06.370 --> 00:09:09.990 align:middle line:84% Debt would have no effect whatsoever. 00:09:09.990 --> 00:09:11.740 align:middle line:84% To get debt to have an effect, though, you 00:09:11.740 --> 00:09:14.630 align:middle line:90% can think about dynamic models. 00:09:14.630 --> 00:09:19.410 align:middle line:84% So one class of model might be if you have a model where 00:09:19.410 --> 00:09:20.610 align:middle line:90% you have customer loyalty. 00:09:20.610 --> 00:09:22.660 align:middle line:84% You sell things to consumers at good prices. 00:09:22.660 --> 00:09:23.920 align:middle line:90% They get used to your product. 00:09:23.920 --> 00:09:25.545 align:middle line:84% They like the taste of your mayonnaise. 00:09:25.545 --> 00:09:28.440 align:middle line:84% They like your frequent flyer miles. 00:09:28.440 --> 00:09:31.210 align:middle line:84% And then you build up this goodwill over time, 00:09:31.210 --> 00:09:33.990 align:middle line:84% but then goodwill is an asset that you can depreciate. 00:09:33.990 --> 00:09:37.110 align:middle line:84% Well, if you now need a lot of money in the short run 00:09:37.110 --> 00:09:39.720 align:middle line:84% to pay down your debt, you might start raising your prices, 00:09:39.720 --> 00:09:42.990 align:middle line:84% exploit all the customers who are temporarily captive to you, 00:09:42.990 --> 00:09:46.650 align:middle line:84% and exploit and burn up all the goodwill you've gotten. 00:09:46.650 --> 00:09:49.320 align:middle line:84% So that might be a case where you raise prices 00:09:49.320 --> 00:09:52.890 align:middle line:84% because you need to raise money in the short run. 00:09:52.890 --> 00:09:55.420 align:middle line:84% In dynamic collusion, you can get the opposite effect. 00:09:55.420 --> 00:09:57.360 align:middle line:84% I talked about in the Rotemberg Saloner model, 00:09:57.360 --> 00:10:00.630 align:middle line:84% if firms have a big incentive to deviate from the cartel 00:10:00.630 --> 00:10:02.550 align:middle line:84% agreement, then the full collusion 00:10:02.550 --> 00:10:04.420 align:middle line:90% is no longer sustainable. 00:10:04.420 --> 00:10:06.870 align:middle line:84% So in a dynamic model, you could have 00:10:06.870 --> 00:10:10.200 align:middle line:84% that we can no longer sustain this high collusive prices 00:10:10.200 --> 00:10:11.585 align:middle line:90% because we have the debt. 00:10:11.585 --> 00:10:13.710 align:middle line:84% We need to lower our prices to reduce our incentive 00:10:13.710 --> 00:10:15.670 align:middle line:90% DV on the cartel agreement. 00:10:15.670 --> 00:10:21.010 align:middle line:84% So this need for short run money could raise or lower prices. 00:10:21.010 --> 00:10:25.300 align:middle line:84% Another effect we talked about with debt is high debt-- 00:10:25.300 --> 00:10:28.390 align:middle line:90% like if there's a-- 00:10:28.390 --> 00:10:29.950 align:middle line:84% for every extra dollar you borrow, 00:10:29.950 --> 00:10:32.283 align:middle line:84% lenders are going to demand a higher and higher interest 00:10:32.283 --> 00:10:35.110 align:middle line:84% rate for the next dollar because the threat of bankruptcy 00:10:35.110 --> 00:10:36.730 align:middle line:90% is larger. 00:10:36.730 --> 00:10:39.910 align:middle line:84% So high debt raises the cost of capital 00:10:39.910 --> 00:10:42.160 align:middle line:84% and makes it difficult for you to borrow more 00:10:42.160 --> 00:10:44.000 align:middle line:90% money for future expansion. 00:10:44.000 --> 00:10:46.330 align:middle line:84% And so if you think about, how would 00:10:46.330 --> 00:10:49.510 align:middle line:84% this work in a standard Cournot competition model, 00:10:49.510 --> 00:10:53.770 align:middle line:84% in a Cournot model, if you've committed to not expand, 00:10:53.770 --> 00:10:55.880 align:middle line:84% then your rivals are going to expand. 00:10:55.880 --> 00:10:58.060 align:middle line:84% Because if I commit to a low Q, my rivals 00:10:58.060 --> 00:11:00.620 align:middle line:84% are going to choose a higher Q as a best response. 00:11:00.620 --> 00:11:04.600 align:middle line:84% So it could be that debt is a commitment that 00:11:04.600 --> 00:11:07.270 align:middle line:90% leads my rivals to expand more. 00:11:07.270 --> 00:11:08.870 align:middle line:84% There could be other models, though. 00:11:08.870 --> 00:11:09.880 align:middle line:90% I haven't talked about them yet. 00:11:09.880 --> 00:11:11.710 align:middle line:84% Like if you have-- there's some new market, 00:11:11.710 --> 00:11:14.380 align:middle line:90% there's some new growing city. 00:11:14.380 --> 00:11:15.940 align:middle line:84% One of two firms is going to open up 00:11:15.940 --> 00:11:17.960 align:middle line:84% a store in that new growing city. 00:11:17.960 --> 00:11:19.030 align:middle line:90% Well, if you know-- 00:11:19.030 --> 00:11:21.338 align:middle line:84% like if you are afraid your rival is going 00:11:21.338 --> 00:11:23.380 align:middle line:84% to open before you, you may race and get in there 00:11:23.380 --> 00:11:24.980 align:middle line:84% before it's short run profitable, 00:11:24.980 --> 00:11:28.040 align:middle line:84% but just when the PDV is positive, to get in 00:11:28.040 --> 00:11:29.810 align:middle line:84% and beat your rival into the new market. 00:11:29.810 --> 00:11:33.530 align:middle line:84% If you know your rival is debt constrained, can't raise money, 00:11:33.530 --> 00:11:36.390 align:middle line:84% you can just wait later and enter later. 00:11:36.390 --> 00:11:39.650 align:middle line:84% So it might be that your rivals enter less if you're 00:11:39.650 --> 00:11:41.930 align:middle line:90% constrained by your debt load. 00:11:41.930 --> 00:11:45.500 align:middle line:84% Anyway, so I tend to think of these as many different stories 00:11:45.500 --> 00:11:49.220 align:middle line:84% in which taking on debt makes firms soft or makes firms tough, 00:11:49.220 --> 00:11:51.980 align:middle line:84% and therefore, would have an entry deterring or entry 00:11:51.980 --> 00:11:54.770 align:middle line:90% accommodating effect. 00:11:54.770 --> 00:11:59.720 align:middle line:84% OK, so this is what Chevalier is interested in, is basically 00:11:59.720 --> 00:12:03.530 align:middle line:84% this question of what effect does debt have, 00:12:03.530 --> 00:12:06.470 align:middle line:84% using the supermarket industry as an opportunity 00:12:06.470 --> 00:12:08.900 align:middle line:90% to explore debt? 00:12:08.900 --> 00:12:11.570 align:middle line:84% And she takes advantage of, again, 00:12:11.570 --> 00:12:13.020 align:middle line:90% what's a natural experiment? 00:12:13.020 --> 00:12:19.110 align:middle line:84% Well, Judy claims that the basic story here, though, 00:12:19.110 --> 00:12:21.870 align:middle line:84% is it's this weird natural experiment in LBOs 00:12:21.870 --> 00:12:25.610 align:middle line:84% in which there's a family called the Haft family. 00:12:25.610 --> 00:12:28.830 align:middle line:84% They had a drugstore chain, the Dart drugstore chain, 00:12:28.830 --> 00:12:30.330 align:middle line:90% and they sold the chain for-- 00:12:30.330 --> 00:12:32.050 align:middle line:84% you know, this is back in the 1990s. 00:12:32.050 --> 00:12:33.925 align:middle line:84% Businesses were worth less than they are now. 00:12:33.925 --> 00:12:36.420 align:middle line:84% They sold their drugstore chain for $160 million. 00:12:36.420 --> 00:12:39.060 align:middle line:84% And soon after, they decided that they 00:12:39.060 --> 00:12:40.380 align:middle line:90% wanted to get into the super-- 00:12:40.380 --> 00:12:42.570 align:middle line:84% now, having cashed out of the drugstore business, 00:12:42.570 --> 00:12:44.070 align:middle line:84% they wanted to get into the supermarket business. 00:12:44.070 --> 00:12:46.050 align:middle line:84% And they embarked on a series of attempts 00:12:46.050 --> 00:12:51.690 align:middle line:84% to buy some national supermarket chain using leveraged buyouts. 00:12:51.690 --> 00:12:56.130 align:middle line:84% So people know what a leveraged buyout is? 00:12:56.130 --> 00:12:58.990 align:middle line:84% Leveraged buyout is, I want to buy something. 00:12:58.990 --> 00:13:00.930 align:middle line:84% I don't have nearly enough money to do it, 00:13:00.930 --> 00:13:03.730 align:middle line:84% but I just get people to lend me a tremendous amount of money. 00:13:03.730 --> 00:13:05.230 align:middle line:84% I take all the borrowed money, and I 00:13:05.230 --> 00:13:08.730 align:middle line:84% use the borrowed money to buy the business that I want to buy. 00:13:08.730 --> 00:13:11.418 align:middle line:84% And so the leverage refers to you're 00:13:11.418 --> 00:13:13.710 align:middle line:84% buying this business largely with someone else's money, 00:13:13.710 --> 00:13:15.627 align:middle line:84% but you're just taking on an awful lot of debt 00:13:15.627 --> 00:13:17.680 align:middle line:90% to buy the business. 00:13:17.680 --> 00:13:21.520 align:middle line:84% Anyway, so first thing we know they did is they tried to buy-- 00:13:21.520 --> 00:13:23.380 align:middle line:84% they just started buying up all the stock 00:13:23.380 --> 00:13:25.600 align:middle line:90% in Safeway on the market. 00:13:25.600 --> 00:13:29.110 align:middle line:84% And then obviously, you start buying up stock, 00:13:29.110 --> 00:13:30.790 align:middle line:84% and you can buy up 5% of the stock 00:13:30.790 --> 00:13:33.130 align:middle line:84% of a publicly traded company before announcing 00:13:33.130 --> 00:13:34.220 align:middle line:90% your intentions. 00:13:34.220 --> 00:13:36.330 align:middle line:84% So they started buying up stock in Safeway. 00:13:36.330 --> 00:13:41.260 align:middle line:84% They then announced, we want to do an LBO and take over Safeway. 00:13:41.260 --> 00:13:42.760 align:middle line:84% Obviously, as soon as they announced 00:13:42.760 --> 00:13:45.250 align:middle line:84% that they were going to take over Safeway, the price of-- 00:13:45.250 --> 00:13:48.940 align:middle line:84% like often you buy out a firm at 50% premium 00:13:48.940 --> 00:13:50.833 align:middle line:84% over the current stock market price. 00:13:50.833 --> 00:13:53.000 align:middle line:84% So they announce they're going to take over Safeway. 00:13:53.000 --> 00:13:57.190 align:middle line:90% Price of Safeway stock jumps. 00:13:57.190 --> 00:14:01.738 align:middle line:84% And then-- so in some sense, they've bought the stock here. 00:14:01.738 --> 00:14:03.280 align:middle line:84% They then announce they're buying it, 00:14:03.280 --> 00:14:05.380 align:middle line:90% and the stock jumps like that. 00:14:05.380 --> 00:14:07.450 align:middle line:84% And then anyway, the Safeway management 00:14:07.450 --> 00:14:10.840 align:middle line:84% does not want to be bought out by the Haft family 00:14:10.840 --> 00:14:12.110 align:middle line:90% and lose their jobs. 00:14:12.110 --> 00:14:14.050 align:middle line:84% So the Safeway management decides, no, we also 00:14:14.050 --> 00:14:15.620 align:middle line:90% think Safeway is undervalued. 00:14:15.620 --> 00:14:18.040 align:middle line:84% We're going to buy Safeway out ourselves. 00:14:18.040 --> 00:14:20.620 align:middle line:84% And so then what the Safeway management did 00:14:20.620 --> 00:14:23.710 align:middle line:84% is they borrowed several billion dollars, 00:14:23.710 --> 00:14:27.250 align:middle line:84% did an LBO to buy the drugstore instead 00:14:27.250 --> 00:14:29.020 align:middle line:90% of letting the Haft family-- 00:14:29.020 --> 00:14:32.080 align:middle line:84% sorry, buy Safeway-- the managers themselves 00:14:32.080 --> 00:14:33.760 align:middle line:84% bought the firm and took it private 00:14:33.760 --> 00:14:39.250 align:middle line:84% rather than letting the Haft family take it over. 00:14:39.250 --> 00:14:42.430 align:middle line:84% And then I think one thing that the Haft family noticed 00:14:42.430 --> 00:14:44.590 align:middle line:84% here was that they made an awful lot of money 00:14:44.590 --> 00:14:49.870 align:middle line:84% when the stock did this, and then after it did this, 00:14:49.870 --> 00:14:54.310 align:middle line:84% then the Safeway management took it over, and then 00:14:54.310 --> 00:14:57.100 align:middle line:84% had to pay them again to buy up all of the shares 00:14:57.100 --> 00:14:59.890 align:middle line:84% that they bought, and to give up their takeover attempt. 00:14:59.890 --> 00:15:05.300 align:middle line:84% And so they ended up making $150 million or something like that. 00:15:05.300 --> 00:15:07.060 align:middle line:84% And then what we don't know is, like, 00:15:07.060 --> 00:15:10.580 align:middle line:84% did the Haft brothers say, wow, that was awesome. 00:15:10.580 --> 00:15:13.680 align:middle line:84% I just made $150 million and I didn't have to do anything, 00:15:13.680 --> 00:15:15.972 align:middle line:84% or did they really want a supermarket chain? 00:15:15.972 --> 00:15:18.430 align:middle line:84% But they basically at this point went from chain, to chain, 00:15:18.430 --> 00:15:21.820 align:middle line:84% to chain, to chain, to chain, and kept trying to buy them, 00:15:21.820 --> 00:15:23.390 align:middle line:90% and kept losing. 00:15:23.390 --> 00:15:27.730 align:middle line:84% And the management of all these different supermarkets 00:15:27.730 --> 00:15:31.132 align:middle line:84% did either LBOs or what are called leveraged recaps where 00:15:31.132 --> 00:15:33.340 align:middle line:84% they just borrowed money, and didn't take it private, 00:15:33.340 --> 00:15:36.580 align:middle line:84% but borrowed a lot of money and bought back a lot of the equity. 00:15:36.580 --> 00:15:39.760 align:middle line:84% And so in this period of just three years, 00:15:39.760 --> 00:15:44.710 align:middle line:84% 19 of the 50 largest supermarket chains in the US did LBOs. 00:15:44.710 --> 00:15:48.790 align:middle line:84% And so Judy treats this as she's going 00:15:48.790 --> 00:15:53.350 align:middle line:84% to think of this as a relatively random choice. 00:15:53.350 --> 00:15:56.233 align:middle line:84% Just where did the Haft brothers go over, and over, 00:15:56.233 --> 00:15:57.650 align:middle line:84% and over again trying to buy them? 00:15:57.650 --> 00:16:00.610 align:middle line:84% And we get this natural experiment of 19 chains 00:16:00.610 --> 00:16:03.420 align:middle line:84% doing LBOs, 31 chains not doing LBOs. 00:16:03.420 --> 00:16:06.310 align:middle line:84% And so you can treat this as a random event. 00:16:06.310 --> 00:16:10.720 align:middle line:84% And then the market level impact of the LBO wave 00:16:10.720 --> 00:16:13.090 align:middle line:84% varied a lot from city to city because most cities 00:16:13.090 --> 00:16:16.830 align:middle line:84% will have three or four large supermarket chains. 00:16:16.830 --> 00:16:18.080 align:middle line:90% And these are national chains. 00:16:18.080 --> 00:16:19.747 align:middle line:84% And so it could be there are some cities 00:16:19.747 --> 00:16:21.997 align:middle line:84% where all the LBO chains had almost the entire market. 00:16:21.997 --> 00:16:23.372 align:middle line:84% There are some cities where there 00:16:23.372 --> 00:16:25.180 align:middle line:84% was only one LBO chain, some cities where 00:16:25.180 --> 00:16:26.970 align:middle line:90% there were no LBO chains. 00:16:26.970 --> 00:16:30.540 align:middle line:90% 00:16:30.540 --> 00:16:32.140 align:middle line:90% OK, so what's her dataset? 00:16:32.140 --> 00:16:34.660 align:middle line:84% It's a few different data sources. 00:16:34.660 --> 00:16:36.450 align:middle line:84% First, she's got annual information 00:16:36.450 --> 00:16:40.200 align:middle line:84% on the number of supermarkets in each chain, each of 85 MSAs. 00:16:40.200 --> 00:16:43.260 align:middle line:90% MSAs means cities. 00:16:43.260 --> 00:16:46.170 align:middle line:84% She has the demographics of the MSAs, 00:16:46.170 --> 00:16:49.230 align:middle line:84% and how that changed over the six-year period, 00:16:49.230 --> 00:16:52.710 align:middle line:84% and then she has whether every store that was there in 1985 00:16:52.710 --> 00:16:59.540 align:middle line:84% became part of an LBO firm by 1988 or by 1991. 00:16:59.540 --> 00:17:02.750 align:middle line:84% So she's going to be running a lot of city-level regressions. 00:17:02.750 --> 00:17:06.319 align:middle line:84% In the city-level regressions, the main explanatory variable 00:17:06.319 --> 00:17:08.510 align:middle line:90% is LBO share. 00:17:08.510 --> 00:17:13.670 align:middle line:84% So LBO share is just the share of stores in 1985 00:17:13.670 --> 00:17:16.022 align:middle line:90% that became part of an LBO. 00:17:16.022 --> 00:17:17.730 align:middle line:84% And so we're going to be comparing cities 00:17:17.730 --> 00:17:19.950 align:middle line:84% that had lots of firms doing LBOs 00:17:19.950 --> 00:17:22.160 align:middle line:90% and not many firms doing LBOs. 00:17:22.160 --> 00:17:23.819 align:middle line:84% And the main thing she's looking at 00:17:23.819 --> 00:17:30.480 align:middle line:84% is interest is the effect of the LBO on entry. 00:17:30.480 --> 00:17:32.280 align:middle line:84% And so she's going to look at both 00:17:32.280 --> 00:17:36.960 align:middle line:84% do the existing non-LBO firms expand and build more stores, 00:17:36.960 --> 00:17:40.030 align:middle line:84% like a Cournot-like model that we can't expand, 00:17:40.030 --> 00:17:42.010 align:middle line:84% therefore they start building more stores, 00:17:42.010 --> 00:17:44.310 align:middle line:84% and then she also looks at effects on entry 00:17:44.310 --> 00:17:48.180 align:middle line:84% by brand new firms come into this market, 00:17:48.180 --> 00:17:51.720 align:middle line:84% perhaps because they know that Safeway can no longer expand, 00:17:51.720 --> 00:17:53.610 align:middle line:84% or they expect Safeway to raise its prices, 00:17:53.610 --> 00:17:54.700 align:middle line:90% or something like that. 00:17:54.700 --> 00:17:57.360 align:middle line:84% So she's going to look at effects on expansions 00:17:57.360 --> 00:17:58.385 align:middle line:90% and on new entry. 00:17:58.385 --> 00:18:02.510 align:middle line:90% 00:18:02.510 --> 00:18:06.320 align:middle line:84% So anyway, it becomes, compared to many papers we've looked at, 00:18:06.320 --> 00:18:08.690 align:middle line:84% fairly simple paper econometrically. 00:18:08.690 --> 00:18:11.750 align:middle line:84% It's just largely a lot of regressions, 00:18:11.750 --> 00:18:16.042 align:middle line:84% looking at how conditions in the market change when-- 00:18:16.042 --> 00:18:18.500 align:middle line:84% differ between the cities that have lots of LBOs and cities 00:18:18.500 --> 00:18:20.090 align:middle line:90% that don't have LBOs. 00:18:20.090 --> 00:18:25.640 align:middle line:84% So the first one here is there is some fairly weak evidence 00:18:25.640 --> 00:18:29.870 align:middle line:84% that number of early LBOs led to an increase 00:18:29.870 --> 00:18:31.860 align:middle line:84% in the total number of stores in the market. 00:18:31.860 --> 00:18:34.700 align:middle line:84% So you can see that that's this coefficient here. 00:18:34.700 --> 00:18:37.400 align:middle line:84% So it's regressing like log of the number of stores 00:18:37.400 --> 00:18:39.720 align:middle line:84% on a market on a bunch of things. 00:18:39.720 --> 00:18:42.020 align:middle line:84% And then one of the things is the-- 00:18:42.020 --> 00:18:51.195 align:middle line:84% and this is log of stores as of 1991 on market characteristics. 00:18:51.195 --> 00:18:52.570 align:middle line:84% And this one here is the fraction 00:18:52.570 --> 00:18:56.920 align:middle line:84% of stores that had done LBOs between 1995 and 1998. 00:18:56.920 --> 00:19:01.870 align:middle line:84% And what it's showing is that in the cities that 00:19:01.870 --> 00:19:07.360 align:middle line:84% have more LBO stores, more stores that did LBOs by 1998, 00:19:07.360 --> 00:19:11.050 align:middle line:84% you see more stores in the market in 2001. 00:19:11.050 --> 00:19:14.680 align:middle line:84% So you would think that's consistent with this story 00:19:14.680 --> 00:19:17.440 align:middle line:84% that LBOs soften price competition. 00:19:17.440 --> 00:19:20.410 align:middle line:84% Safeway after the LBO is going to raise its prices 00:19:20.410 --> 00:19:25.450 align:middle line:84% to try to eat up its customer loyalty and pay off its debts. 00:19:25.450 --> 00:19:27.130 align:middle line:84% Other firms are going to recognize 00:19:27.130 --> 00:19:29.080 align:middle line:84% that as that prices are going to be higher in that market. 00:19:29.080 --> 00:19:31.305 align:middle line:84% We could move in and enter there, or expand there. 00:19:31.305 --> 00:19:37.930 align:middle line:90% 00:19:37.930 --> 00:19:43.000 align:middle line:84% So that was total stores in the market. 00:19:43.000 --> 00:19:50.980 align:middle line:84% So this is a question of when there are more LBO firms, 00:19:50.980 --> 00:19:56.410 align:middle line:84% do the non-LBO firms build more stores in the market? 00:19:56.410 --> 00:19:59.090 align:middle line:90% So this is coefficient here. 00:19:59.090 --> 00:20:03.610 align:middle line:84% If you at the number of stores built by non-LBO incumbents, 00:20:03.610 --> 00:20:09.490 align:middle line:84% when you have an early LBO, the non-LBO incumbents expand more. 00:20:09.490 --> 00:20:12.880 align:middle line:84% Again, it's not highly significant effect, 00:20:12.880 --> 00:20:18.460 align:middle line:84% but non-LBO incumbents expand more 00:20:18.460 --> 00:20:20.850 align:middle line:84% when there are a lot of early LBO firms in the market. 00:20:20.850 --> 00:20:27.010 align:middle line:90% 00:20:27.010 --> 00:20:30.250 align:middle line:84% Third set of regressions is about de novo entry. 00:20:30.250 --> 00:20:34.960 align:middle line:84% So de novo entry is looking at do you see brand new chains that 00:20:34.960 --> 00:20:37.120 align:middle line:84% were never operating in this market before start 00:20:37.120 --> 00:20:38.480 align:middle line:90% to open in stores? 00:20:38.480 --> 00:20:41.320 align:middle line:84% And typically big national chains, they either 00:20:41.320 --> 00:20:43.570 align:middle line:84% have a lot of stores in a city or no stores in a city. 00:20:43.570 --> 00:20:45.612 align:middle line:84% Because if you're going to have stores in a city, 00:20:45.612 --> 00:20:47.450 align:middle line:84% you need warehouses and things like that. 00:20:47.450 --> 00:20:49.420 align:middle line:84% And then occasionally, the big chains 00:20:49.420 --> 00:20:51.820 align:middle line:84% will decide, OK, now we're moving into this metro area, 00:20:51.820 --> 00:20:53.920 align:middle line:90% and they'll build out there. 00:20:53.920 --> 00:20:55.750 align:middle line:84% Also at the time, you were getting 00:20:55.750 --> 00:20:59.410 align:middle line:84% Kmart and Walmart were expanding from just other things 00:20:59.410 --> 00:21:01.690 align:middle line:90% into groceries. 00:21:01.690 --> 00:21:04.480 align:middle line:90% And so what they find on-- 00:21:04.480 --> 00:21:07.150 align:middle line:84% what she finds on de novo entry is, 00:21:07.150 --> 00:21:09.050 align:middle line:90% again, this coefficient here. 00:21:09.050 --> 00:21:13.478 align:middle line:84% The more early LBOs there are in a city, the more likely 00:21:13.478 --> 00:21:15.520 align:middle line:84% it is that we're going to see some new chain that 00:21:15.520 --> 00:21:18.670 align:middle line:84% wasn't in the city, or Kmart, or Walmart opening up 00:21:18.670 --> 00:21:22.170 align:middle line:90% new supermarkets there. 00:21:22.170 --> 00:21:25.650 align:middle line:84% So I guess, overall, these results 00:21:25.650 --> 00:21:31.830 align:middle line:84% are consistent with models in which LBOs make incumbents soft. 00:21:31.830 --> 00:21:34.760 align:middle line:84% This could be-- what the soft is in terms of entry, 00:21:34.760 --> 00:21:37.350 align:middle line:84% it could be the LBO firms are raising prices, 00:21:37.350 --> 00:21:40.230 align:middle line:84% or it could be that the LBO firms are capacity constrained. 00:21:40.230 --> 00:21:43.530 align:middle line:84% And that leads the rivals to introduce more stores. 00:21:43.530 --> 00:21:46.710 align:middle line:84% Certainly, if it's the raising prices part, 00:21:46.710 --> 00:21:50.070 align:middle line:84% the raising prices one would be, that is-- 00:21:50.070 --> 00:21:53.310 align:middle line:84% you do want to be soft in price competition games. 00:21:53.310 --> 00:21:55.770 align:middle line:84% So if it's a commitment to raise your prices that 00:21:55.770 --> 00:21:59.100 align:middle line:84% gets your rivals to raise their prices, that's good. 00:21:59.100 --> 00:22:02.910 align:middle line:84% This isn't to say that you do more-- 00:22:02.910 --> 00:22:04.950 align:middle line:84% this isn't to say that strategic effect of LBOs 00:22:04.950 --> 00:22:06.585 align:middle line:84% is what makes you do an LBO, but it 00:22:06.585 --> 00:22:08.460 align:middle line:84% is saying that it's one of the considerations 00:22:08.460 --> 00:22:10.140 align:middle line:84% in whether to take an LBO is-- one 00:22:10.140 --> 00:22:12.240 align:middle line:84% of the benefits the LBO would be providing 00:22:12.240 --> 00:22:14.670 align:middle line:84% is the committing to raise your price, which 00:22:14.670 --> 00:22:16.295 align:middle line:84% gets your rivals to raise their prices. 00:22:16.295 --> 00:22:18.628 align:middle line:84% And that would be one of the factors in the calculation. 00:22:18.628 --> 00:22:20.550 align:middle line:84% So in some sense, you would do more LBOs 00:22:20.550 --> 00:22:23.850 align:middle line:84% than you would in the open loop LBO equilibrium 00:22:23.850 --> 00:22:26.430 align:middle line:84% where the rivals didn't know that you were doing the LBO 00:22:26.430 --> 00:22:27.950 align:middle line:84% and it couldn't affect your pricing. 00:22:27.950 --> 00:22:30.880 align:middle line:90% 00:22:30.880 --> 00:22:33.000 align:middle line:90% Questions on that? 00:22:33.000 --> 00:22:36.990 align:middle line:84% And then Judy also has a nice follow-up paper 00:22:36.990 --> 00:22:41.340 align:middle line:84% in the Journal of Finance which looks directly at pricing. 00:22:41.340 --> 00:22:48.010 align:middle line:84% So this one, again, it's a fairly-- 00:22:48.010 --> 00:22:50.435 align:middle line:84% she's gotten data from supermarket scanner data, 00:22:50.435 --> 00:22:52.060 align:middle line:84% and looking at supermarket scanner data 00:22:52.060 --> 00:22:54.580 align:middle line:84% to compare how the firms with undertaken LBOs 00:22:54.580 --> 00:22:56.750 align:middle line:84% priced compared to non-LBO firms. 00:22:56.750 --> 00:22:59.500 align:middle line:84% And what she finds is that by 1992, 00:22:59.500 --> 00:23:04.610 align:middle line:84% the firms that did LBOs have raised their prices by 2% to 3%. 00:23:04.610 --> 00:23:08.380 align:middle line:90% 00:23:08.380 --> 00:23:12.790 align:middle line:84% In markets where there isn't some dominant firm that 00:23:12.790 --> 00:23:16.480 align:middle line:84% didn't do an LBO, then market level prices 00:23:16.480 --> 00:23:18.770 align:middle line:90% typically go up after the LBO. 00:23:18.770 --> 00:23:24.580 align:middle line:84% But in markets where you do have the dominant low debt firm, what 00:23:24.580 --> 00:23:27.370 align:middle line:84% we see is that actually, on average in the market prices, 00:23:27.370 --> 00:23:29.510 align:middle line:84% if the market level prices go down, 00:23:29.510 --> 00:23:33.070 align:middle line:84% it seems like sometimes the low debt firms 00:23:33.070 --> 00:23:35.140 align:middle line:90% are exploiting the LBO firms. 00:23:35.140 --> 00:23:37.600 align:middle line:84% And the LBO firms are committing to cut prices, 00:23:37.600 --> 00:23:39.190 align:middle line:84% and the LBO firms perhaps could be 00:23:39.190 --> 00:23:42.520 align:middle line:84% deciding this is an opportunity for us to keep our prices low. 00:23:42.520 --> 00:23:46.490 align:middle line:84% And we see some increased exits of LBO stores 00:23:46.490 --> 00:23:48.830 align:middle line:84% where the LBO stores find it difficult to compete 00:23:48.830 --> 00:23:50.410 align:middle line:90% with the non-LBO firms. 00:23:50.410 --> 00:23:54.270 align:middle line:90% 00:23:54.270 --> 00:23:55.440 align:middle line:90% Questions? 00:23:55.440 --> 00:23:57.900 align:middle line:84% I mean, it's two nice papers that 00:23:57.900 --> 00:24:00.062 align:middle line:84% are very much of that first approach 00:24:00.062 --> 00:24:01.770 align:middle line:84% I said of just looking at the investment, 00:24:01.770 --> 00:24:03.210 align:middle line:84% seeing what the effect of the investment is. 00:24:03.210 --> 00:24:04.252 align:middle line:90% What does it do to firms? 00:24:04.252 --> 00:24:07.810 align:middle line:90% 00:24:07.810 --> 00:24:12.050 align:middle line:84% OK, so then the second paper I wanted to talk about, 00:24:12.050 --> 00:24:18.340 align:middle line:84% this is a paper Sara and I wrote about entity deterrence 00:24:18.340 --> 00:24:20.680 align:middle line:84% in the pharmaceutical industries. 00:24:20.680 --> 00:24:25.000 align:middle line:84% So I guess basics of pharmaceuticals 00:24:25.000 --> 00:24:30.010 align:middle line:90% is, you invent a new drug. 00:24:30.010 --> 00:24:31.430 align:middle line:90% You patent the new drug. 00:24:31.430 --> 00:24:32.830 align:middle line:84% After you patent the drug, you do 00:24:32.830 --> 00:24:34.480 align:middle line:84% some number of years of clinical trials 00:24:34.480 --> 00:24:37.825 align:middle line:84% to prove that the drug is safe and effective. 00:24:37.825 --> 00:24:39.700 align:middle line:84% If you have enough, you present your evidence 00:24:39.700 --> 00:24:42.260 align:middle line:84% that drug is safe and effective to the FDA. 00:24:42.260 --> 00:24:46.300 align:middle line:84% They approve your drug for sale in the United States. 00:24:46.300 --> 00:24:49.510 align:middle line:84% The clock has been ticking since you 00:24:49.510 --> 00:24:51.610 align:middle line:84% patented the drug on how long you will 00:24:51.610 --> 00:24:53.480 align:middle line:90% have patent protection over it. 00:24:53.480 --> 00:24:55.720 align:middle line:84% There are various more complicated rules 00:24:55.720 --> 00:24:59.590 align:middle line:84% about if it's taken you this long to get the drug approved, 00:24:59.590 --> 00:25:01.650 align:middle line:84% you can get extensions to your patent term. 00:25:01.650 --> 00:25:04.150 align:middle line:84% And there are also various ways you can get other extensions 00:25:04.150 --> 00:25:04.983 align:middle line:90% to your patent term. 00:25:04.983 --> 00:25:07.345 align:middle line:84% But all drugs will have some date at which-- 00:25:07.345 --> 00:25:10.840 align:middle line:90% 00:25:10.840 --> 00:25:14.590 align:middle line:84% they'll have some date at which the price drops. 00:25:14.590 --> 00:25:18.520 align:middle line:84% Prices of many drugs are quite high. 00:25:18.520 --> 00:25:21.560 align:middle line:84% I think these days, most people have insurance. 00:25:21.560 --> 00:25:24.790 align:middle line:84% But because most people have insurance, 00:25:24.790 --> 00:25:27.482 align:middle line:84% many new drugs come out that treat high blood pressure, 00:25:27.482 --> 00:25:28.940 align:middle line:84% or cholesterol reduction, whatever, 00:25:28.940 --> 00:25:31.690 align:middle line:84% and they come out with prices of $1,000 a month, 00:25:31.690 --> 00:25:34.730 align:middle line:90% or $2,000 a month, or whatever. 00:25:34.730 --> 00:25:38.570 align:middle line:84% And then obviously, you have insurance as a copay. 00:25:38.570 --> 00:25:41.860 align:middle line:84% But the firms are getting these very high $2,000 a month prices 00:25:41.860 --> 00:25:42.650 align:middle line:90% for new drugs. 00:25:42.650 --> 00:25:44.775 align:middle line:84% Now, sometimes people are looking at pricing things 00:25:44.775 --> 00:25:46.870 align:middle line:90% at $50,000 a month. 00:25:46.870 --> 00:25:49.720 align:middle line:84% At some point, the patent expires. 00:25:49.720 --> 00:25:53.620 align:middle line:84% When the patent expires, there's something 00:25:53.620 --> 00:25:56.498 align:middle line:84% called an abbreviated new drug approval, ANDA. 00:25:56.498 --> 00:25:57.915 align:middle line:84% I think that's what it stands for. 00:25:57.915 --> 00:26:00.610 align:middle line:90% 00:26:00.610 --> 00:26:03.700 align:middle line:84% Rivals can come in, show that they're 00:26:03.700 --> 00:26:05.320 align:middle line:84% producing a chemical compound that's 00:26:05.320 --> 00:26:07.570 align:middle line:84% equivalent to the chemical compound being produced 00:26:07.570 --> 00:26:12.670 align:middle line:84% by the branded drug manufacturer that was approved by the FDA. 00:26:12.670 --> 00:26:15.010 align:middle line:84% Earlier in US history, you had to again prove 00:26:15.010 --> 00:26:17.135 align:middle line:84% safety and effectiveness, and there was essentially 00:26:17.135 --> 00:26:17.870 align:middle line:90% no generic entry. 00:26:17.870 --> 00:26:22.120 align:middle line:84% Now, all you have to do is prove bioequivalence of your drug 00:26:22.120 --> 00:26:23.540 align:middle line:90% with the existing drug. 00:26:23.540 --> 00:26:27.310 align:middle line:84% Then you can come in and compete with the existing firm. 00:26:27.310 --> 00:26:30.130 align:middle line:84% Typically, what happens is firms are going along charging 00:26:30.130 --> 00:26:34.450 align:middle line:84% their $1,000 a month price, and then know, generics enter. 00:26:34.450 --> 00:26:36.962 align:middle line:90% And generics just come in and-- 00:26:36.962 --> 00:26:38.920 align:middle line:84% there's some date the generic entry is allowed. 00:26:38.920 --> 00:26:41.410 align:middle line:84% And the generic just goes like that and prices 00:26:41.410 --> 00:26:44.150 align:middle line:84% at a small fraction of what the branded firm charges. 00:26:44.150 --> 00:26:47.725 align:middle line:84% Strangely, the branded firm does not cut its price to match. 00:26:47.725 --> 00:26:50.350 align:middle line:84% Typically, what the branded firm does, if anything, is do this, 00:26:50.350 --> 00:26:52.300 align:middle line:90% and price even higher. 00:26:52.300 --> 00:26:56.800 align:middle line:84% Because now, all they're serving is the people who will not buy-- 00:26:56.800 --> 00:27:00.340 align:middle line:84% people who will not buy a bioequivalent drug, and so 00:27:00.340 --> 00:27:02.260 align:middle line:84% those people you might think are captive 00:27:02.260 --> 00:27:04.680 align:middle line:84% and are not going to leave no matter what you charge. 00:27:04.680 --> 00:27:06.730 align:middle line:84% But in any case, if you look at the market 00:27:06.730 --> 00:27:09.982 align:middle line:84% shares of the branded drug, the market shares go like this. 00:27:09.982 --> 00:27:11.690 align:middle line:84% The branded drug has a market share of 1, 00:27:11.690 --> 00:27:14.950 align:middle line:84% and then it pretty quickly asymptotes to some very small 00:27:14.950 --> 00:27:16.220 align:middle line:90% number, like 10%. 00:27:16.220 --> 00:27:18.280 align:middle line:84% And then the generics go like this, 00:27:18.280 --> 00:27:21.480 align:middle line:84% and then they start to survey the whole market. 00:27:21.480 --> 00:27:28.720 align:middle line:84% So what's nice about drugs as a place 00:27:28.720 --> 00:27:31.810 align:middle line:84% to study strategic investment is there's 00:27:31.810 --> 00:27:34.280 align:middle line:84% no entry possible until some date, 00:27:34.280 --> 00:27:36.530 align:middle line:84% and then there's some date on which entry is possible. 00:27:36.530 --> 00:27:39.460 align:middle line:84% And looking at the FDA data, you know the date 00:27:39.460 --> 00:27:41.630 align:middle line:90% at which the entry is possible. 00:27:41.630 --> 00:27:44.770 align:middle line:84% And so we get this opportunity where there's no entry, 00:27:44.770 --> 00:27:46.400 align:middle line:84% no entry possible, no entry possible. 00:27:46.400 --> 00:27:48.130 align:middle line:84% Suddenly entry is possible, and you 00:27:48.130 --> 00:27:51.610 align:middle line:84% can use that change in how firms behave around the time entry 00:27:51.610 --> 00:27:54.040 align:middle line:84% becomes possible to identify what firms 00:27:54.040 --> 00:27:55.395 align:middle line:90% might be doing to deter entry. 00:27:55.395 --> 00:27:58.330 align:middle line:90% 00:27:58.330 --> 00:28:01.000 align:middle line:84% So anyway, our data sample in this paper, 00:28:01.000 --> 00:28:04.960 align:middle line:84% it's 63 branded drugs that lost patent protection at some point 00:28:04.960 --> 00:28:11.160 align:middle line:90% between 1986 and 1992. 00:28:11.160 --> 00:28:13.360 align:middle line:90% I think I said that already. 00:28:13.360 --> 00:28:19.280 align:middle line:84% So this paper is, as I said, the reduced form approach. 00:28:19.280 --> 00:28:22.780 align:middle line:84% We're trying to understand not what the incentives are. 00:28:22.780 --> 00:28:26.380 align:middle line:84% We think that in many cases, the economics is simple enough 00:28:26.380 --> 00:28:29.320 align:middle line:84% that we know what the incentives for pharmaceutical manufacturers 00:28:29.320 --> 00:28:32.770 align:middle line:84% are to do, and how they would deter entry. 00:28:32.770 --> 00:28:35.980 align:middle line:84% The question is, are they reacting to those incentives 00:28:35.980 --> 00:28:39.070 align:middle line:84% to deter entry, and do we see them taking actions 00:28:39.070 --> 00:28:42.370 align:middle line:84% to try to keep generic drugs out of their markets? 00:28:42.370 --> 00:28:44.320 align:middle line:84% And the way we propose to do this 00:28:44.320 --> 00:28:47.987 align:middle line:84% is to identify a prediction that we think differs 00:28:47.987 --> 00:28:50.320 align:middle line:84% between the strategic model and the non-strategic model, 00:28:50.320 --> 00:28:52.420 align:middle line:90% and then test that prediction. 00:28:52.420 --> 00:28:56.320 align:middle line:84% So the prediction comes off two observations. 00:28:56.320 --> 00:29:00.190 align:middle line:84% First observation is that strategic entry deterrence 00:29:00.190 --> 00:29:05.170 align:middle line:84% models often predict that the relationship between investment 00:29:05.170 --> 00:29:08.890 align:middle line:84% and market size is going to be non monotonic. 00:29:08.890 --> 00:29:16.550 align:middle line:84% The idea is that if you're Eli Lilly and your patent on Prozac 00:29:16.550 --> 00:29:20.240 align:middle line:84% is expiring, or if you're Pfizer and your patent on Lipitor 00:29:20.240 --> 00:29:22.820 align:middle line:84% is expiring, there's nothing you can do 00:29:22.820 --> 00:29:24.650 align:middle line:90% to stop entry from occurring. 00:29:24.650 --> 00:29:28.460 align:middle line:84% Because Lipitor is bringing in billions of dollars 00:29:28.460 --> 00:29:29.430 align:middle line:90% a year in profits. 00:29:29.430 --> 00:29:31.920 align:middle line:90% It costs almost nothing to make. 00:29:31.920 --> 00:29:33.810 align:middle line:84% As soon as the patent on Lipitor expires, 00:29:33.810 --> 00:29:37.970 align:middle line:84% there's going to be a ton of generic atorvastatin firms. 00:29:37.970 --> 00:29:41.670 align:middle line:84% There's nothing that Pfizer could do about that. 00:29:41.670 --> 00:29:43.820 align:middle line:90% Most drugs in the data-- 00:29:43.820 --> 00:29:46.320 align:middle line:84% we think of drugs as having these high prices and incredibly 00:29:46.320 --> 00:29:47.257 align:middle line:90% high revenues. 00:29:47.257 --> 00:29:48.840 align:middle line:84% If you actually look at the full cross 00:29:48.840 --> 00:29:51.810 align:middle line:84% section of FDA approved drugs with expiring patents, 00:29:51.810 --> 00:29:54.000 align:middle line:90% most of them don't do anything. 00:29:54.000 --> 00:29:56.380 align:middle line:90% Firms get drugs. 00:29:56.380 --> 00:29:58.150 align:middle line:90% They get the drugs approved. 00:29:58.150 --> 00:30:00.760 align:middle line:84% But then the drug gets approved, but it's not 00:30:00.760 --> 00:30:03.190 align:middle line:84% as good as some other drug in the market. 00:30:03.190 --> 00:30:06.710 align:middle line:84% Or it was the best drug when it was first patented, 00:30:06.710 --> 00:30:09.220 align:middle line:84% but then some new drug has come in and supplanted it. 00:30:09.220 --> 00:30:12.040 align:middle line:84% So many, many drugs, basically, to first order approximation, 00:30:12.040 --> 00:30:13.300 align:middle line:90% nobody takes. 00:30:13.300 --> 00:30:15.010 align:middle line:84% And if you have a drug that nobody takes, 00:30:15.010 --> 00:30:16.802 align:middle line:84% there's never going to be any generic entry 00:30:16.802 --> 00:30:20.770 align:middle line:84% because the generic firm is not going to come into a market that 00:30:20.770 --> 00:30:22.600 align:middle line:84% already has no patients when it's got 00:30:22.600 --> 00:30:24.410 align:middle line:90% pharmaceutical reps pushing it. 00:30:24.410 --> 00:30:28.210 align:middle line:84% And so our thought on the paper is that these entry deterrence 00:30:28.210 --> 00:30:30.700 align:middle line:84% incentives really are something that exist just 00:30:30.700 --> 00:30:32.560 align:middle line:84% in intermediate sized markets, and markets 00:30:32.560 --> 00:30:35.850 align:middle line:84% that are big enough so that generic firms would want 00:30:35.850 --> 00:30:38.970 align:middle line:84% to bother, but small enough that it's plausible that you could 00:30:38.970 --> 00:30:40.710 align:middle line:84% keep the generic firms out by taking 00:30:40.710 --> 00:30:42.430 align:middle line:90% some set of strategic actions. 00:30:42.430 --> 00:30:46.530 align:middle line:84% And so there's this potential non-monotonic effect of there 00:30:46.530 --> 00:30:48.130 align:middle line:90% will be some action. 00:30:48.130 --> 00:30:50.670 align:middle line:84% And in small markets, you don't do this action. 00:30:50.670 --> 00:30:52.540 align:middle line:84% And in big markets, you do this action. 00:30:52.540 --> 00:30:56.470 align:middle line:84% And then in really big markets, you don't do the action again. 00:30:56.470 --> 00:31:01.230 align:middle line:84% And then the other observation is that non-strategic models-- 00:31:01.230 --> 00:31:05.910 align:middle line:84% so if I'm putting here market size on this, 00:31:05.910 --> 00:31:10.420 align:middle line:84% and investment on this axis, the other observation 00:31:10.420 --> 00:31:15.070 align:middle line:84% is that non-strategic models like the model ignoring 00:31:15.070 --> 00:31:17.350 align:middle line:84% that entry deterrence incentives often 00:31:17.350 --> 00:31:19.660 align:middle line:84% predict that the relationship between market size 00:31:19.660 --> 00:31:21.940 align:middle line:84% and investment is going to be monotone. 00:31:21.940 --> 00:31:25.690 align:middle line:84% We talk in the paper about there being basically two effects that 00:31:25.690 --> 00:31:29.080 align:middle line:84% determine whether we should see more or less investments 00:31:29.080 --> 00:31:31.300 align:middle line:90% in big markets. 00:31:31.300 --> 00:31:33.865 align:middle line:84% One is the direct effect, and it's just, 00:31:33.865 --> 00:31:38.530 align:middle line:84% is a marginal investment return larger in a big market 00:31:38.530 --> 00:31:39.980 align:middle line:90% than it is in a small market? 00:31:39.980 --> 00:31:42.640 align:middle line:84% And many investments like TV commercials, 00:31:42.640 --> 00:31:46.900 align:middle line:84% like you may be aware that there's just-- if you watch TV, 00:31:46.900 --> 00:31:49.120 align:middle line:84% there's zillions of these annoying commercials that 00:31:49.120 --> 00:31:51.295 align:middle line:84% go on for a minute talking about all the conditions 00:31:51.295 --> 00:31:53.920 align:middle line:84% that someone might have, and all the side effects of this drug, 00:31:53.920 --> 00:31:59.320 align:middle line:84% and buy Xeljanz or whatever at the end of the commercial. 00:31:59.320 --> 00:32:02.477 align:middle line:84% Obviously, commercials like that are only 00:32:02.477 --> 00:32:04.810 align:middle line:84% going to be feasible for conditions that a lot of people 00:32:04.810 --> 00:32:05.418 align:middle line:90% have. 00:32:05.418 --> 00:32:07.960 align:middle line:84% If there's some condition that's very rare in the population, 00:32:07.960 --> 00:32:10.100 align:middle line:84% you can't buy a national TV commercial 00:32:10.100 --> 00:32:11.195 align:middle line:90% advertising your drug. 00:32:11.195 --> 00:32:13.070 align:middle line:84% If there's a condition that many people have, 00:32:13.070 --> 00:32:14.420 align:middle line:90% you can buy the commercial. 00:32:14.420 --> 00:32:18.855 align:middle line:84% So we'd call that the direct effect of investment returns. 00:32:18.855 --> 00:32:20.480 align:middle line:84% The more people who have the condition, 00:32:20.480 --> 00:32:25.720 align:middle line:84% the more you're going to invest in over-the-air TV advertising. 00:32:25.720 --> 00:32:30.430 align:middle line:84% A second effect-- do I have second effect here?-- 00:32:30.430 --> 00:32:32.470 align:middle line:90% strategic effect. 00:32:32.470 --> 00:32:35.860 align:middle line:84% The second strategic effect is, are the returns 00:32:35.860 --> 00:32:38.500 align:middle line:84% to this investment bigger in a monopoly market, 00:32:38.500 --> 00:32:40.240 align:middle line:90% or in a duopoly market? 00:32:40.240 --> 00:32:45.970 align:middle line:84% Because the bigger is the market, the more likely 00:32:45.970 --> 00:32:48.108 align:middle line:84% you are to end up with competition. 00:32:48.108 --> 00:32:49.900 align:middle line:84% And so you're going to have to think about, 00:32:49.900 --> 00:32:52.390 align:middle line:84% is this an investment that I'm making more valuable when 00:32:52.390 --> 00:32:54.040 align:middle line:84% there's competition, or more value when 00:32:54.040 --> 00:32:55.780 align:middle line:90% there's no competition? 00:32:55.780 --> 00:32:58.817 align:middle line:84% So if those two effects go in the same way-- 00:32:58.817 --> 00:33:01.150 align:middle line:84% this is something you want to invest more in big markets 00:33:01.150 --> 00:33:02.590 align:middle line:84% than small markets, and this thing 00:33:02.590 --> 00:33:04.990 align:middle line:84% you want to invest more in oligopoly markets 00:33:04.990 --> 00:33:06.520 align:middle line:90% than in monopoly markets-- 00:33:06.520 --> 00:33:08.800 align:middle line:84% then we're going to see a monotone relationship 00:33:08.800 --> 00:33:11.170 align:middle line:84% between market size and the investment. 00:33:11.170 --> 00:33:13.570 align:middle line:84% And so if you have investments like that, 00:33:13.570 --> 00:33:15.500 align:middle line:84% we think those two effects go in the same way, 00:33:15.500 --> 00:33:16.875 align:middle line:84% we would expect it to be monotone 00:33:16.875 --> 00:33:18.808 align:middle line:84% if there wasn't the entry deterrence motive. 00:33:18.808 --> 00:33:20.350 align:middle line:84% But with the entry deterrence motive, 00:33:20.350 --> 00:33:24.700 align:middle line:84% we could look for a non-monotone pattern. 00:33:24.700 --> 00:33:27.770 align:middle line:84% I think I have several more slides on this. 00:33:27.770 --> 00:33:29.570 align:middle line:90% I'll see how many of them I do. 00:33:29.570 --> 00:33:33.920 align:middle line:84% So here would be an example of a simple numerical model 00:33:33.920 --> 00:33:40.490 align:middle line:84% I can solve that has this property, monotone effect 00:33:40.490 --> 00:33:41.960 align:middle line:84% without entry deterrence motives, 00:33:41.960 --> 00:33:44.460 align:middle line:84% and a non-monotone with the entry deterrence motives. 00:33:44.460 --> 00:33:47.540 align:middle line:84% So here, think about a cross section of markets, 00:33:47.540 --> 00:33:51.020 align:middle line:84% and the markets are identical other than the number of people, 00:33:51.020 --> 00:33:52.520 align:middle line:90% zi, who-- 00:33:52.520 --> 00:33:55.220 align:middle line:84% you can think of this as the market for a particular drug, 00:33:55.220 --> 00:34:00.250 align:middle line:84% and zi is the number of people who would be served by the drug. 00:34:00.250 --> 00:34:03.640 align:middle line:84% Assume that each market has a continuum 00:34:03.640 --> 00:34:08.199 align:middle line:84% of types with theta distributed uniformly on 0 and 1. 00:34:08.199 --> 00:34:13.280 align:middle line:84% Assume that this advertising has this technology where 00:34:13.280 --> 00:34:18.440 align:middle line:84% A is my per-consumer spending on the advertising. 00:34:18.440 --> 00:34:22.190 align:middle line:84% And if I spend A per consumer on the advertising, 00:34:22.190 --> 00:34:27.250 align:middle line:84% then consumers are going to get utility theta times root 2A 00:34:27.250 --> 00:34:30.429 align:middle line:84% minus p1 if they buy the branded drug, 00:34:30.429 --> 00:34:34.530 align:middle line:84% 1/2 theta root 2A minus p if they buy the generic drug, 00:34:34.530 --> 00:34:38.380 align:middle line:84% my rival, and 0 if they buy neither good. 00:34:38.380 --> 00:34:40.150 align:middle line:84% So a motivation for this technology 00:34:40.150 --> 00:34:46.060 align:middle line:84% would be some drugs treat diseases that are typically 00:34:46.060 --> 00:34:48.760 align:middle line:90% managed by specialists. 00:34:48.760 --> 00:34:51.850 align:middle line:84% The way you market drugs that are sold to specialists, rather 00:34:51.850 --> 00:34:54.940 align:middle line:84% than putting on TV commercials, you just send sales reps 00:34:54.940 --> 00:34:56.060 align:middle line:90% to the specialists. 00:34:56.060 --> 00:34:58.480 align:middle line:84% So if this is a not very common disease, 00:34:58.480 --> 00:35:02.050 align:middle line:84% you have the top 25 specialists in this disease in the Boston 00:35:02.050 --> 00:35:02.620 align:middle line:90% area. 00:35:02.620 --> 00:35:05.390 align:middle line:84% You send a sales rep to visit every one of them. 00:35:05.390 --> 00:35:08.200 align:middle line:84% You tell them one on one about the drug and all the newest 00:35:08.200 --> 00:35:10.600 align:middle line:84% data on its effectiveness, and you try 00:35:10.600 --> 00:35:12.400 align:middle line:90% to get them to prescribe it. 00:35:12.400 --> 00:35:15.970 align:middle line:84% So we're modeling that as the more you spend on advertising, 00:35:15.970 --> 00:35:18.740 align:middle line:84% the higher utility people will have for buying your drug. 00:35:18.740 --> 00:35:21.340 align:middle line:84% But because the doctors you're talking to 00:35:21.340 --> 00:35:23.920 align:middle line:84% will know that the generic is also 00:35:23.920 --> 00:35:26.350 align:middle line:84% much the same in this model, I'm going 00:35:26.350 --> 00:35:28.100 align:middle line:84% to assume that people get utility-- 00:35:28.100 --> 00:35:29.860 align:middle line:84% when you raise people's willingness 00:35:29.860 --> 00:35:32.120 align:middle line:84% to pay for your drug, you also raise their willingness 00:35:32.120 --> 00:35:34.250 align:middle line:90% to pay for the generic drug. 00:35:34.250 --> 00:35:38.104 align:middle line:90% What happens in this model-- 00:35:38.104 --> 00:35:38.812 align:middle line:90% give me a second. 00:35:38.812 --> 00:35:39.771 align:middle line:90% It's right here or not. 00:35:39.771 --> 00:35:40.430 align:middle line:90% Nope, OK. 00:35:40.430 --> 00:35:42.020 align:middle line:84% Anyway, what happens in this model 00:35:42.020 --> 00:35:44.000 align:middle line:84% I will explain it a little more later. 00:35:44.000 --> 00:35:47.210 align:middle line:84% If there's no strategic entry deterrence motive, 00:35:47.210 --> 00:35:52.390 align:middle line:84% this is a model where you would do less investing the larger is 00:35:52.390 --> 00:35:52.905 align:middle line:90% the market. 00:35:52.905 --> 00:35:55.030 align:middle line:84% And you're going to do less investing the larger is 00:35:55.030 --> 00:35:58.330 align:middle line:84% the market because the larger is the market, the more likely 00:35:58.330 --> 00:36:00.670 align:middle line:84% it is that the generic competitor is 00:36:00.670 --> 00:36:02.180 align:middle line:90% going to end up entering. 00:36:02.180 --> 00:36:04.360 align:middle line:84% If the generic competitor ends up entering, 00:36:04.360 --> 00:36:08.230 align:middle line:84% it's as if the gain you're getting is 1/2 theta root 2A 00:36:08.230 --> 00:36:10.750 align:middle line:90% instead of theta root 2A. 00:36:10.750 --> 00:36:13.263 align:middle line:84% Because it's only the marginal way 00:36:13.263 --> 00:36:15.430 align:middle line:84% that increases your value relative to the competitor 00:36:15.430 --> 00:36:16.130 align:middle line:90% that matters. 00:36:16.130 --> 00:36:19.300 align:middle line:84% So the bigger is the market, the more likely generics 00:36:19.300 --> 00:36:24.260 align:middle line:90% are going to enter and profit. 00:36:24.260 --> 00:36:26.140 align:middle line:84% And so therefore, advertising is monotone 00:36:26.140 --> 00:36:29.308 align:middle line:90% decreasing in the market size. 00:36:29.308 --> 00:36:30.850 align:middle line:84% In some sense, this is what you would 00:36:30.850 --> 00:36:33.490 align:middle line:84% do if you knew entry was never going to occur, 00:36:33.490 --> 00:36:35.865 align:middle line:84% and this is what you would do if you knew entry was going 00:36:35.865 --> 00:36:38.420 align:middle line:90% to occur with probability 1. 00:36:38.420 --> 00:36:41.030 align:middle line:84% But if you look at the strategic entry in this model, 00:36:41.030 --> 00:36:42.620 align:middle line:84% the strategic entry will be like this. 00:36:42.620 --> 00:36:45.140 align:middle line:84% It'll have this non-monotone pattern where 00:36:45.140 --> 00:36:48.800 align:middle line:84% it departs from the strategic one, dips way down, 00:36:48.800 --> 00:36:50.420 align:middle line:90% and then comes back up. 00:36:50.420 --> 00:36:52.130 align:middle line:90% What's this dip? 00:36:52.130 --> 00:36:56.690 align:middle line:84% The dip is I'm trying to keep my market small enough 00:36:56.690 --> 00:36:59.520 align:middle line:84% so that the generics don't come in and compete with me. 00:36:59.520 --> 00:37:02.030 align:middle line:84% So what I do is I do less advertising. 00:37:02.030 --> 00:37:04.910 align:middle line:84% With less advertising, I have lower sales. 00:37:04.910 --> 00:37:07.922 align:middle line:84% Lower sales, less likely there's a generic competitor. 00:37:07.922 --> 00:37:09.380 align:middle line:84% And so what I'm doing in some sense 00:37:09.380 --> 00:37:11.870 align:middle line:84% is I'm keeping the market small so 00:37:11.870 --> 00:37:14.540 align:middle line:84% that I can get monopoly profits in the small market 00:37:14.540 --> 00:37:18.370 align:middle line:84% instead of getting oligopoly profits in a bigger market. 00:37:18.370 --> 00:37:21.660 align:middle line:90% 00:37:21.660 --> 00:37:23.140 align:middle line:90% And at some point, I give up. 00:37:23.140 --> 00:37:25.902 align:middle line:84% Here, I start increasing my advertising again 00:37:25.902 --> 00:37:27.610 align:middle line:84% because I realize that generics are going 00:37:27.610 --> 00:37:28.693 align:middle line:90% to come in no matter what. 00:37:28.693 --> 00:37:31.600 align:middle line:84% So I might as well be splitting the best possible market 00:37:31.600 --> 00:37:33.670 align:middle line:84% with the generic competitor rather than 00:37:33.670 --> 00:37:36.340 align:middle line:84% trying to keep it small enough to keep the generics out. 00:37:36.340 --> 00:37:40.930 align:middle line:84% Anyway, so the example above, I was comparing investment 00:37:40.930 --> 00:37:42.500 align:middle line:90% in two different models. 00:37:42.500 --> 00:37:45.490 align:middle line:84% The upper line that I drew for you, 00:37:45.490 --> 00:37:49.330 align:middle line:84% that was the line for this no entry deterrence motive model 00:37:49.330 --> 00:37:52.780 align:middle line:84% where the potential entrant only observes the advertising level 00:37:52.780 --> 00:37:56.080 align:middle line:84% A after it's come into the market or not. 00:37:56.080 --> 00:37:59.860 align:middle line:84% And then the other line was the standard strategic entry 00:37:59.860 --> 00:38:03.145 align:middle line:84% deterrence model where I choose A, they observe A, 00:38:03.145 --> 00:38:04.780 align:middle line:90% and they can stay out. 00:38:04.780 --> 00:38:10.803 align:middle line:84% And so our argument was that the relationship between market 00:38:10.803 --> 00:38:11.303 align:middle line:90% size-- 00:38:11.303 --> 00:38:15.240 align:middle line:90% 00:38:15.240 --> 00:38:17.220 align:middle line:84% determining market size in the investment, A, 00:38:17.220 --> 00:38:22.890 align:middle line:84% was going to be monotone on the bottom model and non-monotone 00:38:22.890 --> 00:38:26.280 align:middle line:90% in the model on the top. 00:38:26.280 --> 00:38:29.110 align:middle line:84% So anyway, the paper tries to explain 00:38:29.110 --> 00:38:33.880 align:middle line:84% why is it that we're going to see these monotone patterns 00:38:33.880 --> 00:38:36.690 align:middle line:90% versus non-monotone patterns? 00:38:36.690 --> 00:38:39.240 align:middle line:84% How do we think about-- the strategic entry deterrence 00:38:39.240 --> 00:38:43.790 align:middle line:84% model, you have a profit function pi 1 of A. 00:38:43.790 --> 00:38:48.320 align:middle line:84% And I guess I didn't say here, you know, 00:38:48.320 --> 00:38:51.710 align:middle line:84% in all of Tirole's models, Tirole always 00:38:51.710 --> 00:38:54.710 align:middle line:84% has the entry cost, E being known. 00:38:54.710 --> 00:38:56.570 align:middle line:84% And the entry cost, E, being known 00:38:56.570 --> 00:38:59.030 align:middle line:84% means if I want to keep the entrant out, 00:38:59.030 --> 00:39:00.450 align:middle line:90% I have to distort this far. 00:39:00.450 --> 00:39:02.867 align:middle line:84% And going this far keeps them out, or going less than that 00:39:02.867 --> 00:39:03.960 align:middle line:90% doesn't keep them out. 00:39:03.960 --> 00:39:06.170 align:middle line:84% In empirical work, we're always going 00:39:06.170 --> 00:39:08.600 align:middle line:84% to be making the entry cost, E, random 00:39:08.600 --> 00:39:13.820 align:middle line:84% and have it unpredictable to the incumbent exactly what's going 00:39:13.820 --> 00:39:17.020 align:middle line:90% to keep the entrant out or not. 00:39:17.020 --> 00:39:23.570 align:middle line:84% So if I think about what's my profit if I choose investment A, 00:39:23.570 --> 00:39:25.070 align:middle line:90% I don't know the entry cost. 00:39:25.070 --> 00:39:29.630 align:middle line:84% So assume that the entry cost, E, has some distribution, F. 00:39:29.630 --> 00:39:35.270 align:middle line:84% So F of pi 2 duopoly of A is the probability that my rival's 00:39:35.270 --> 00:39:38.600 align:middle line:84% entry cost draw is low enough that the rival comes 00:39:38.600 --> 00:39:41.120 align:middle line:84% in and competes with me, in which case 00:39:41.120 --> 00:39:47.080 align:middle line:84% I will get the duopoly profits that you get after investment A. 00:39:47.080 --> 00:39:50.420 align:middle line:84% The other case is when the entry cost is high. 00:39:50.420 --> 00:39:53.950 align:middle line:84% If the entry cost is high, which happens with 1 minus F of pi 2D 00:39:53.950 --> 00:39:58.030 align:middle line:84% star of A, then I'm going to get to be a monopolist and get this 00:39:58.030 --> 00:39:59.380 align:middle line:90% profit level. 00:39:59.380 --> 00:40:02.210 align:middle line:84% And then I have a cost of doing the advertising. 00:40:02.210 --> 00:40:06.910 align:middle line:90% 00:40:06.910 --> 00:40:11.330 align:middle line:84% The model on the bottom of my board 00:40:11.330 --> 00:40:13.250 align:middle line:84% was different in that the entrant only 00:40:13.250 --> 00:40:16.370 align:middle line:84% observed A after making its entry decision. 00:40:16.370 --> 00:40:18.290 align:middle line:84% So again, this is a shorthand for your not 00:40:18.290 --> 00:40:22.460 align:middle line:84% thinking about the effect of A on the entrant. 00:40:22.460 --> 00:40:27.740 align:middle line:84% So if the entrant wasn't observing A, 00:40:27.740 --> 00:40:29.750 align:middle line:90% what advertising would you do? 00:40:29.750 --> 00:40:32.000 align:middle line:84% There, the entry decision doesn't 00:40:32.000 --> 00:40:35.700 align:middle line:84% depend on A. The entry deterrent depends on A star-- 00:40:35.700 --> 00:40:40.500 align:middle line:84% what the incumbent expects you to-- what 00:40:40.500 --> 00:40:42.783 align:middle line:84% the potential entrant expects you to invest. 00:40:42.783 --> 00:40:44.700 align:middle line:84% So you want to think about writing your profit 00:40:44.700 --> 00:40:46.290 align:middle line:90% as a function of two things-- 00:40:46.290 --> 00:40:48.870 align:middle line:84% the actual advertising that you do, 00:40:48.870 --> 00:40:51.166 align:middle line:84% and the equilibrium level of advertising. 00:40:51.166 --> 00:40:53.984 align:middle line:90% 00:40:53.984 --> 00:40:59.120 align:middle line:84% And here, the firm enters if its entry cost 00:40:59.120 --> 00:41:00.980 align:middle line:84% is less than what duopoly profits would 00:41:00.980 --> 00:41:04.040 align:middle line:84% be at the equilibrium level of advertising. 00:41:04.040 --> 00:41:06.020 align:middle line:84% But then, the actual profits received 00:41:06.020 --> 00:41:08.490 align:middle line:84% depend on the true level of advertising. 00:41:08.490 --> 00:41:10.790 align:middle line:84% Again, entry depends on the equilibrium level 00:41:10.790 --> 00:41:12.180 align:middle line:90% of advertising. 00:41:12.180 --> 00:41:15.270 align:middle line:84% Monopoly profit depends on the true advertising. 00:41:15.270 --> 00:41:17.960 align:middle line:84% Cost of it depends on the true advertising. 00:41:17.960 --> 00:41:21.470 align:middle line:84% So when I think about the first order conditions for this profit 00:41:21.470 --> 00:41:23.600 align:middle line:84% maximization problem versus this one, 00:41:23.600 --> 00:41:25.250 align:middle line:84% the two first order conditions are 00:41:25.250 --> 00:41:28.130 align:middle line:84% going to differ because of the fact 00:41:28.130 --> 00:41:31.800 align:middle line:84% that here, there's a derivative with respect to A in this term, 00:41:31.800 --> 00:41:35.120 align:middle line:84% and here, changing A doesn't affect this term. 00:41:35.120 --> 00:41:38.160 align:middle line:84% So that's what makes the two problems different. 00:41:38.160 --> 00:41:40.750 align:middle line:90% 00:41:40.750 --> 00:41:43.010 align:middle line:90% What is the extra term? 00:41:43.010 --> 00:41:45.610 align:middle line:84% So they have the same first order conditions. 00:41:45.610 --> 00:41:48.340 align:middle line:84% Except if you think about what's the effect that 00:41:48.340 --> 00:41:50.710 align:middle line:84% comes from differentiating this with respect to A, 00:41:50.710 --> 00:41:55.060 align:middle line:84% it's that it's the derivative of this with respect to A times 00:41:55.060 --> 00:42:00.340 align:middle line:84% this plus the derivative of this with respect to A times that. 00:42:00.340 --> 00:42:02.030 align:middle line:90% That's the new term. 00:42:02.030 --> 00:42:05.830 align:middle line:84% So the extra entry deterrence term is-- 00:42:05.830 --> 00:42:09.360 align:middle line:90% and this is intuitive-- 00:42:09.360 --> 00:42:14.850 align:middle line:84% how much does A affect the duopoly profits 00:42:14.850 --> 00:42:20.980 align:middle line:84% that the firm is going to get times the probability 00:42:20.980 --> 00:42:24.130 align:middle line:84% that that effect on profits is going 00:42:24.130 --> 00:42:26.560 align:middle line:90% to affect 2's entry decision. 00:42:26.560 --> 00:42:28.990 align:middle line:84% And it's going to affect 2's entry decision if and only 00:42:28.990 --> 00:42:33.610 align:middle line:84% if his entry cost is identical to the profits, 00:42:33.610 --> 00:42:36.410 align:middle line:90% or within D pi of the profits. 00:42:36.410 --> 00:42:42.617 align:middle line:84% And then if you do effect entry, the extra profit that you get 00:42:42.617 --> 00:42:44.200 align:middle line:84% is the difference between the monopoly 00:42:44.200 --> 00:42:45.660 align:middle line:90% profit and the duopoly profit. 00:42:45.660 --> 00:42:49.004 align:middle line:90% 00:42:49.004 --> 00:42:50.600 align:middle line:90% And this term-- 00:42:50.600 --> 00:42:52.570 align:middle line:90% [INAUDIBLE] says that? 00:42:52.570 --> 00:42:53.070 align:middle line:90% No. 00:42:53.070 --> 00:42:55.360 align:middle line:90% OK, yeah. 00:42:55.360 --> 00:42:59.770 align:middle line:84% This term is where the non-monotonicity comes from, 00:42:59.770 --> 00:43:03.190 align:middle line:84% because this term is multiplied by the likelihood 00:43:03.190 --> 00:43:05.620 align:middle line:84% that a small decrease in an entrant's profits 00:43:05.620 --> 00:43:08.900 align:middle line:84% is going to be pivotal to their entry decision. 00:43:08.900 --> 00:43:11.092 align:middle line:84% And so this is what creates the non-monotonicity. 00:43:11.092 --> 00:43:13.550 align:middle line:84% You know, this we think of as a big number between monopoly 00:43:13.550 --> 00:43:15.120 align:middle line:90% and duopoly profits. 00:43:15.120 --> 00:43:17.690 align:middle line:84% This is just whatever effect your advertising 00:43:17.690 --> 00:43:18.663 align:middle line:90% has on the rival. 00:43:18.663 --> 00:43:20.330 align:middle line:84% But the thing that makes it non-monotone 00:43:20.330 --> 00:43:23.130 align:middle line:84% is, is that change in your rival's profit 00:43:23.130 --> 00:43:25.310 align:middle line:84% is going to be pivotal to whether they enter or not? 00:43:25.310 --> 00:43:27.180 align:middle line:84% And that's going to be non-monotone. 00:43:27.180 --> 00:43:31.836 align:middle line:90% 00:43:31.836 --> 00:43:35.740 align:middle line:84% And then the second effect, I said this earlier in the paper, 00:43:35.740 --> 00:43:39.910 align:middle line:84% there are these two reasons why investment normally 00:43:39.910 --> 00:43:41.770 align:middle line:90% changes in market size. 00:43:41.770 --> 00:43:46.430 align:middle line:84% It changes for this direct effect which I define here. 00:43:46.430 --> 00:43:50.650 align:middle line:84% So there's this direct effect, which is just 00:43:50.650 --> 00:43:52.640 align:middle line:90% its direct effect on pi 2. 00:43:52.640 --> 00:43:56.500 align:middle line:84% And then there's the competition effect. 00:43:56.500 --> 00:43:58.390 align:middle line:84% So the competition effect is, do you 00:43:58.390 --> 00:44:06.520 align:middle line:84% want to do more in a duopoly or a monopoly? 00:44:06.520 --> 00:44:10.600 align:middle line:84% And the direct effect is, do you want 00:44:10.600 --> 00:44:14.050 align:middle line:84% to do more in a big market, or a small market? 00:44:14.050 --> 00:44:15.940 align:middle line:84% And so in some sense, the theorem 00:44:15.940 --> 00:44:18.760 align:middle line:84% says that there are, in that model, exactly two reasons 00:44:18.760 --> 00:44:22.630 align:middle line:84% why you're changing investment as market size increases. 00:44:22.630 --> 00:44:24.130 align:middle line:84% One is because of the direct effect, 00:44:24.130 --> 00:44:26.130 align:middle line:84% and the other is because the competition effect. 00:44:26.130 --> 00:44:28.870 align:middle line:84% If those two go in the same direction, 00:44:28.870 --> 00:44:33.415 align:middle line:84% then the investment would be monotone. 00:44:33.415 --> 00:44:35.540 align:middle line:84% And we can think about whether we expect investment 00:44:35.540 --> 00:44:37.290 align:middle line:84% to be monotone by just thinking about what 00:44:37.290 --> 00:44:41.300 align:middle line:84% are direct and competition effects of an investment. 00:44:41.300 --> 00:44:46.460 align:middle line:84% And what I'd done in this model is clearly, 00:44:46.460 --> 00:44:50.300 align:middle line:84% the fact that the advertising raised the value for the generic 00:44:50.300 --> 00:44:52.800 align:middle line:84% is what gave it a negative competition effect. 00:44:52.800 --> 00:44:54.800 align:middle line:84% If my rival is going to be benefiting from that 00:44:54.800 --> 00:44:57.200 align:middle line:90% and raising-- 00:44:57.200 --> 00:45:03.320 align:middle line:84% I'm getting a smaller value in a duopoly, 00:45:03.320 --> 00:45:06.530 align:middle line:84% I eliminated the direct effect by saying 00:45:06.530 --> 00:45:10.650 align:middle line:84% A is a per-consumer spending on advertising. 00:45:10.650 --> 00:45:12.110 align:middle line:84% And so it's a per-consumer spending 00:45:12.110 --> 00:45:14.420 align:middle line:84% for a per-consumer benefit, so there's 00:45:14.420 --> 00:45:16.790 align:middle line:84% no direct effect, big versus small, 00:45:16.790 --> 00:45:19.670 align:middle line:84% if you're thinking of this as per-consumer spending. 00:45:19.670 --> 00:45:22.190 align:middle line:84% And that technology where you're visiting doctors, 00:45:22.190 --> 00:45:24.260 align:middle line:84% and the number of doctors you have to visit 00:45:24.260 --> 00:45:26.060 align:middle line:84% is proportional number of patients, 00:45:26.060 --> 00:45:29.060 align:middle line:84% it's plausible that the per-consumer cost 00:45:29.060 --> 00:45:31.100 align:middle line:84% and the per-consumer benefit are the same. 00:45:31.100 --> 00:45:34.320 align:middle line:84% So I've just basically eliminated direct effect. 00:45:34.320 --> 00:45:36.050 align:middle line:84% And so once you eliminate direct effects, 00:45:36.050 --> 00:45:37.550 align:middle line:84% it's going to be monotone increasing 00:45:37.550 --> 00:45:39.380 align:middle line:84% or monotone decreasing depending on whether the competition 00:45:39.380 --> 00:45:41.280 align:middle line:90% effect is positive or negative. 00:45:41.280 --> 00:45:44.300 align:middle line:90% 00:45:44.300 --> 00:45:45.870 align:middle line:90% Anyway, paper does other stuff. 00:45:45.870 --> 00:45:51.890 align:middle line:84% So let me go to empirical application here. 00:45:51.890 --> 00:45:55.140 align:middle line:84% There are 63 branded drugs in our data sample. 00:45:55.140 --> 00:45:58.340 align:middle line:84% All the branded drugs lost patent protection 00:45:58.340 --> 00:46:02.060 align:middle line:84% at some point between 1986 and 1992. 00:46:02.060 --> 00:46:05.720 align:middle line:84% And so we're going to think about 00:46:05.720 --> 00:46:07.610 align:middle line:84% how does the behavior of incumbents 00:46:07.610 --> 00:46:09.663 align:middle line:84% compare between big and small markets? 00:46:09.663 --> 00:46:14.000 align:middle line:90% 00:46:14.000 --> 00:46:15.950 align:middle line:90% So we're going to put-- 00:46:15.950 --> 00:46:17.960 align:middle line:84% generally, you're going to make up 00:46:17.960 --> 00:46:22.250 align:middle line:84% graphs where we have market size on the x-axis. 00:46:22.250 --> 00:46:29.310 align:middle line:84% What market size means for us is revenue some number of years 00:46:29.310 --> 00:46:30.655 align:middle line:90% prior to patent expiration. 00:46:30.655 --> 00:46:33.030 align:middle line:84% The thought is that if you're going to be doing something 00:46:33.030 --> 00:46:35.370 align:middle line:84% with your drug to deter entry, that 00:46:35.370 --> 00:46:37.155 align:middle line:90% comes closer to the entry date. 00:46:37.155 --> 00:46:39.780 align:middle line:84% And so we're going to look a few years back from the entry date 00:46:39.780 --> 00:46:41.520 align:middle line:90% and say, how big was the market? 00:46:41.520 --> 00:46:45.330 align:middle line:84% That's our possibly exogenous measure of market size. 00:46:45.330 --> 00:46:48.750 align:middle line:84% And then we're going to look at for potential strategic 00:46:48.750 --> 00:46:50.490 align:middle line:90% investments. 00:46:50.490 --> 00:46:52.540 align:middle line:90% One is detail advertising. 00:46:52.540 --> 00:46:54.720 align:middle line:90% So detail advertising is-- 00:46:54.720 --> 00:46:56.730 align:middle line:84% strangely enough, data on this exists-- 00:46:56.730 --> 00:46:59.250 align:middle line:84% the expenditures by pharmaceutical firms 00:46:59.250 --> 00:47:02.010 align:middle line:84% sending sales representatives to doctors offices 00:47:02.010 --> 00:47:03.880 align:middle line:90% to tell them about their drugs. 00:47:03.880 --> 00:47:06.515 align:middle line:84% And we have data on that at the drug level. 00:47:06.515 --> 00:47:09.330 align:middle line:90% 00:47:09.330 --> 00:47:14.920 align:middle line:84% Second strategic investment we have is journal advertising. 00:47:14.920 --> 00:47:18.840 align:middle line:84% This is-- one thing you do with drugs is advertise them 00:47:18.840 --> 00:47:21.550 align:middle line:84% in medical journals that are sold to doctors. 00:47:21.550 --> 00:47:24.240 align:middle line:84% And so we know, for each drug in our sample, 00:47:24.240 --> 00:47:27.300 align:middle line:84% how much money they were spending advertising them 00:47:27.300 --> 00:47:30.400 align:middle line:90% to physicians. 00:47:30.400 --> 00:47:33.070 align:middle line:84% Third strategic investment we look at 00:47:33.070 --> 00:47:35.920 align:middle line:90% is presentation proliferation. 00:47:35.920 --> 00:47:38.410 align:middle line:84% Something that people have discussed 00:47:38.410 --> 00:47:42.400 align:middle line:84% firms doing to deter entry is, if you 00:47:42.400 --> 00:47:46.550 align:middle line:84% go into your doctor with a prescription for-- 00:47:46.550 --> 00:47:48.550 align:middle line:84% you go into the pharmacy and have a prescription 00:47:48.550 --> 00:47:54.250 align:middle line:84% for a 50-milligram tablet of some medicine, 00:47:54.250 --> 00:47:56.960 align:middle line:84% and then they don't have a 50 milligram tablet, 00:47:56.960 --> 00:47:59.530 align:middle line:84% they have a 100 milligram tablet. 00:47:59.530 --> 00:48:02.255 align:middle line:84% Sometimes, they can give you the 100 milligram tablets, 00:48:02.255 --> 00:48:03.880 align:middle line:84% and give you a little pill cutter thing 00:48:03.880 --> 00:48:05.150 align:middle line:90% that cuts the pills in half. 00:48:05.150 --> 00:48:06.940 align:middle line:84% And you can turn one pill into two, 00:48:06.940 --> 00:48:09.850 align:middle line:84% and take half of the 100 milligram pill, 00:48:09.850 --> 00:48:11.570 align:middle line:90% but sometimes you can't. 00:48:11.570 --> 00:48:13.810 align:middle line:84% And certainly, if they've given you 00:48:13.810 --> 00:48:17.415 align:middle line:84% a prescription for an ointment that contains this drug, 00:48:17.415 --> 00:48:18.790 align:middle line:84% and they don't have the ointment, 00:48:18.790 --> 00:48:20.680 align:middle line:84% they only have the oral, the pharmacist 00:48:20.680 --> 00:48:23.728 align:middle line:84% can't sell you the generic ointment when you're 00:48:23.728 --> 00:48:25.020 align:middle line:90% supposed to be taking the pill. 00:48:25.020 --> 00:48:26.250 align:middle line:84% They can't sell you the gel cap if you're 00:48:26.250 --> 00:48:27.840 align:middle line:84% supposed to be taking the pill, or the liquid 00:48:27.840 --> 00:48:29.740 align:middle line:84% if you're supposed to be taking the pill, or whatever. 00:48:29.740 --> 00:48:32.282 align:middle line:84% And so something you've heard that pharmaceutical firms might 00:48:32.282 --> 00:48:35.610 align:middle line:84% do to deter entry is start producing 00:48:35.610 --> 00:48:38.940 align:middle line:84% the drug in a zillion different varieties to, in some sense, 00:48:38.940 --> 00:48:42.760 align:middle line:84% raise the cost of the entrance to come in and compete with you. 00:48:42.760 --> 00:48:45.660 align:middle line:84% So if, instead of having one standard pill, 00:48:45.660 --> 00:48:46.590 align:middle line:90% I have a whole bunch-- 00:48:46.590 --> 00:48:48.548 align:middle line:84% I have a regular pill, and a time release pill, 00:48:48.548 --> 00:48:50.340 align:middle line:84% and a gel cap, and a something or other-- 00:48:50.340 --> 00:48:53.565 align:middle line:90% I can have many different-- 00:48:53.565 --> 00:48:55.440 align:middle line:84% if the generic only came in and only produced 00:48:55.440 --> 00:48:57.610 align:middle line:84% the most popular presentation of the drug, 00:48:57.610 --> 00:49:00.027 align:middle line:84% it would only be competing for a small part of the market. 00:49:00.027 --> 00:49:02.520 align:middle line:84% And that could make it more difficult to enter. 00:49:02.520 --> 00:49:06.280 align:middle line:84% And then the final thing that we do is just prices. 00:49:06.280 --> 00:49:11.870 align:middle line:84% And again, I discussed some of these issues 00:49:11.870 --> 00:49:14.870 align:middle line:84% about the entry deterrence-- the entry deterrence 00:49:14.870 --> 00:49:17.660 align:middle line:84% through signaling costs, or signaling something 00:49:17.660 --> 00:49:20.120 align:middle line:84% about demand, or something where you set low prices 00:49:20.120 --> 00:49:21.990 align:middle line:90% to try to deter entry. 00:49:21.990 --> 00:49:25.040 align:middle line:84% And so we're going to look at these investments 00:49:25.040 --> 00:49:28.850 align:middle line:84% and how the investments look when you graph them 00:49:28.850 --> 00:49:29.837 align:middle line:90% against market size. 00:49:29.837 --> 00:49:30.920 align:middle line:90% And do they look monotone? 00:49:30.920 --> 00:49:32.520 align:middle line:90% Do they look non-monotone? 00:49:32.520 --> 00:49:34.250 align:middle line:84% And what we're going to particularly do 00:49:34.250 --> 00:49:38.930 align:middle line:84% is focus on four drugs that are in that intermediate range 00:49:38.930 --> 00:49:40.910 align:middle line:90% where entry is uncertain. 00:49:40.910 --> 00:49:43.100 align:middle line:84% Is that where we see a distortion in the direction 00:49:43.100 --> 00:49:46.430 align:middle line:84% that you would expect the distortion? 00:49:46.430 --> 00:49:49.280 align:middle line:84% And what distortions would you expect? 00:49:49.280 --> 00:49:51.200 align:middle line:84% If I'm trying to deter entry, I'm 00:49:51.200 --> 00:49:53.480 align:middle line:84% going to do less detail advertising because I'm trying 00:49:53.480 --> 00:49:55.130 align:middle line:90% to keep the market small. 00:49:55.130 --> 00:49:57.710 align:middle line:84% I'm going to do less journal advertising because I'm trying 00:49:57.710 --> 00:49:59.310 align:middle line:90% to keep the market small. 00:49:59.310 --> 00:50:01.790 align:middle line:84% I'm going to do more presentation proliferation 00:50:01.790 --> 00:50:04.910 align:middle line:84% and introduce more versions of my drug, 00:50:04.910 --> 00:50:07.760 align:middle line:84% and I'm going to set lower prices. 00:50:07.760 --> 00:50:09.990 align:middle line:84% And I'm going to look for are they 00:50:09.990 --> 00:50:13.140 align:middle line:84% non-monotone with those deviations from monotonicity 00:50:13.140 --> 00:50:15.060 align:middle line:84% in the range where it's plausible 00:50:15.060 --> 00:50:18.750 align:middle line:90% that you could keep rivals out? 00:50:18.750 --> 00:50:21.580 align:middle line:84% And we're going to look at this in two different ways. 00:50:21.580 --> 00:50:24.330 align:middle line:84% One, just in the pure cross section, 00:50:24.330 --> 00:50:27.180 align:middle line:84% advertising versus market size, is advertising 00:50:27.180 --> 00:50:29.280 align:middle line:90% non-monotone in revenue? 00:50:29.280 --> 00:50:32.353 align:middle line:84% But sometimes, these things are just super steeply 00:50:32.353 --> 00:50:34.020 align:middle line:84% increasing depending on how you describe 00:50:34.020 --> 00:50:36.187 align:middle line:84% the variables, because just bigger markets-- like TV 00:50:36.187 --> 00:50:37.710 align:middle line:84% advertising, this would never work 00:50:37.710 --> 00:50:39.990 align:middle line:84% because TV advertising, the bigger the market is, 00:50:39.990 --> 00:50:42.090 align:middle line:90% the more you spend. 00:50:42.090 --> 00:50:44.760 align:middle line:84% So the other thing we do is look at how 00:50:44.760 --> 00:50:47.588 align:middle line:84% are firms changing advertising in the year 00:50:47.588 --> 00:50:49.380 align:middle line:84% the patent's expiring relative to what they 00:50:49.380 --> 00:50:51.980 align:middle line:84% were doing two years before the patent expires? 00:50:51.980 --> 00:50:54.780 align:middle line:84% And so you can think of it as it's like we're using AT minus 2 00:50:54.780 --> 00:50:57.510 align:middle line:84% as the control level of the investment, 00:50:57.510 --> 00:51:00.360 align:middle line:84% and think about how are they changing investments? 00:51:00.360 --> 00:51:05.895 align:middle line:84% And is the change in investment non-monotone in that kind 00:51:05.895 --> 00:51:06.395 align:middle line:90% of pattern? 00:51:06.395 --> 00:51:10.570 align:middle line:90% 00:51:10.570 --> 00:51:12.420 align:middle line:84% And what are some of the findings? 00:51:12.420 --> 00:51:15.940 align:middle line:84% And again, we're very conservative in this paper 00:51:15.940 --> 00:51:19.450 align:middle line:84% in looking for general tests of non-monotonicity, 00:51:19.450 --> 00:51:21.520 align:middle line:84% non-parametrically, rather than just putting 00:51:21.520 --> 00:51:22.910 align:middle line:90% in quadratic functions. 00:51:22.910 --> 00:51:25.160 align:middle line:84% And so nothing's all that significant. 00:51:25.160 --> 00:51:27.610 align:middle line:84% But there are a few places where it looks like we 00:51:27.610 --> 00:51:29.350 align:middle line:90% find these non-monotonicities. 00:51:29.350 --> 00:51:33.770 align:middle line:84% The first is in the detail in journal advertising. 00:51:33.770 --> 00:51:38.950 align:middle line:84% When you just graph them versus the market size, 00:51:38.950 --> 00:51:41.440 align:middle line:84% it looks like the detail advertising-- 00:51:41.440 --> 00:51:42.770 align:middle line:90% these are just the means-- 00:51:42.770 --> 00:51:44.830 align:middle line:84% the detail advertising is relatively 00:51:44.830 --> 00:51:47.290 align:middle line:84% low for the second quintile drugs 00:51:47.290 --> 00:51:49.630 align:middle line:90% relative to the other drugs. 00:51:49.630 --> 00:51:52.480 align:middle line:84% And the journal advertising is relatively 00:51:52.480 --> 00:51:54.610 align:middle line:84% low for the second quintile drugs 00:51:54.610 --> 00:51:56.720 align:middle line:90% relative to the other drugs. 00:51:56.720 --> 00:51:59.500 align:middle line:84% So it looks like there's this cross-sectional pattern 00:51:59.500 --> 00:52:01.540 align:middle line:84% in the detail in journal advertising 00:52:01.540 --> 00:52:03.820 align:middle line:84% where in these second quintile drugs where 00:52:03.820 --> 00:52:07.616 align:middle line:84% the entry is uncertain, there is lower levels of advertising. 00:52:07.616 --> 00:52:10.600 align:middle line:90% 00:52:10.600 --> 00:52:16.480 align:middle line:84% But marginally significant once you test it non-parametrically. 00:52:16.480 --> 00:52:23.290 align:middle line:84% And then the other thing we look for is 00:52:23.290 --> 00:52:28.000 align:middle line:84% how are things changing as patent expiration approaches? 00:52:28.000 --> 00:52:32.620 align:middle line:84% And here, we get some marginally significant results 00:52:32.620 --> 00:52:35.080 align:middle line:84% on the presentation proliferation 00:52:35.080 --> 00:52:38.200 align:middle line:90% and on the journal advertising. 00:52:38.200 --> 00:52:41.590 align:middle line:84% The journal advertising is going in the wrong direction, 00:52:41.590 --> 00:52:43.660 align:middle line:84% but the presentation Herfindahl, there's 00:52:43.660 --> 00:52:46.900 align:middle line:84% some evidence of the presentation proliferation 00:52:46.900 --> 00:52:50.182 align:middle line:84% also going up for these second quintile drugs. 00:52:50.182 --> 00:52:51.640 align:middle line:84% This looks like they're introducing 00:52:51.640 --> 00:52:54.070 align:middle line:84% new versions of the second quintile drugs 00:52:54.070 --> 00:52:56.890 align:middle line:84% as the patent is expiring that they hadn't introduced years 00:52:56.890 --> 00:52:58.430 align:middle line:84% before when they were monopolist. 00:52:58.430 --> 00:53:03.860 align:middle line:90% 00:53:03.860 --> 00:53:06.010 align:middle line:90% Questions? 00:53:06.010 --> 00:53:06.680 align:middle line:90% Comments? 00:53:06.680 --> 00:53:08.840 align:middle line:90% No? 00:53:08.840 --> 00:53:10.980 align:middle line:90% OK. 00:53:10.980 --> 00:53:15.650 align:middle line:84% So the third paper, this one I'm going to do super quickly. 00:53:15.650 --> 00:53:22.250 align:middle line:84% Goolsbee-Syverson is-- it's another reduced form prediction 00:53:22.250 --> 00:53:24.680 align:middle line:84% paper looking at some prediction that 00:53:24.680 --> 00:53:29.000 align:middle line:84% would hold for strategic models versus non-strategic models. 00:53:29.000 --> 00:53:34.790 align:middle line:84% The prediction that they explore is, how do firms change prices 00:53:34.790 --> 00:53:38.010 align:middle line:84% when they become threatened by Southwest's entry 00:53:38.010 --> 00:53:42.236 align:middle line:84% even though Southwest has not entered? 00:53:42.236 --> 00:53:45.860 align:middle line:84% And so what's really nice about this application is, 00:53:45.860 --> 00:53:49.670 align:middle line:90% imagine you have a market-- 00:53:49.670 --> 00:53:52.550 align:middle line:84% so we have a market that Southwest is not currently 00:53:52.550 --> 00:54:00.350 align:middle line:84% serving, like flying from Boston to Nashville. 00:54:00.350 --> 00:54:03.990 align:middle line:84% And there are no flights on Southwest 00:54:03.990 --> 00:54:05.670 align:middle line:90% from Boston to Nashville. 00:54:05.670 --> 00:54:09.720 align:middle line:84% And the current carriers serving Boston to Nashville 00:54:09.720 --> 00:54:13.680 align:middle line:84% are completely unscared of Southwest entering and competing 00:54:13.680 --> 00:54:15.690 align:middle line:84% with them because Southwest doesn't have 00:54:15.690 --> 00:54:16.968 align:middle line:90% any operations in Nashville. 00:54:16.968 --> 00:54:18.010 align:middle line:90% They have no gates there. 00:54:18.010 --> 00:54:19.177 align:middle line:90% They have no aircraft there. 00:54:19.177 --> 00:54:20.530 align:middle line:90% They have no employees there. 00:54:20.530 --> 00:54:22.982 align:middle line:84% If Southwest doesn't have any employees in Nashville, 00:54:22.982 --> 00:54:24.690 align:middle line:84% it's unlikely they're going to just start 00:54:24.690 --> 00:54:27.534 align:middle line:90% a Boston to Nashville flight. 00:54:27.534 --> 00:54:32.680 align:middle line:84% But suppose Southwest does operate out of Boston. 00:54:32.680 --> 00:54:34.710 align:middle line:84% Suppose Southwest were to announce, 00:54:34.710 --> 00:54:36.870 align:middle line:84% we're moving into Nashville, and we're initially 00:54:36.870 --> 00:54:40.710 align:middle line:84% going to fly Nashville to Miami, Nashville to Dallas, Nashville 00:54:40.710 --> 00:54:45.390 align:middle line:90% to Las Vegas, Nashville to LA. 00:54:45.390 --> 00:54:48.390 align:middle line:84% Well, then, if you're the carriers on the Boston 00:54:48.390 --> 00:54:51.930 align:middle line:84% to Nashville route, you become scared 00:54:51.930 --> 00:54:56.640 align:middle line:84% that Southwest, now that they have operations in both Boston 00:54:56.640 --> 00:54:59.040 align:middle line:84% and Nashville, are they going to start serving 00:54:59.040 --> 00:55:01.200 align:middle line:90% the Boston to Nashville market? 00:55:01.200 --> 00:55:04.950 align:middle line:84% But the nice feature that Goolsbee and Syverson note 00:55:04.950 --> 00:55:09.030 align:middle line:84% is that Southwest both announces that they're 00:55:09.030 --> 00:55:12.810 align:middle line:84% coming to Nashville five months before they come to Nashville, 00:55:12.810 --> 00:55:14.610 align:middle line:84% and when they announce we're going 00:55:14.610 --> 00:55:16.420 align:middle line:84% to come to Nashville with these routes, 00:55:16.420 --> 00:55:19.440 align:middle line:84% so you know they're not serving Boston to Nashville. 00:55:19.440 --> 00:55:22.320 align:middle line:84% And so what Goolsbee and Syverson say is, 00:55:22.320 --> 00:55:25.740 align:middle line:84% let's imagine that firms are trying to keep-- 00:55:25.740 --> 00:55:31.410 align:middle line:84% that once this happens, suddenly United Airlines 00:55:31.410 --> 00:55:35.820 align:middle line:84% has an incentive to try to keep Southwest out 00:55:35.820 --> 00:55:37.210 align:middle line:90% of the Boston Nashville market. 00:55:37.210 --> 00:55:38.390 align:middle line:90% What might they do? 00:55:38.390 --> 00:55:41.460 align:middle line:84% They're looking at, again, this limit pricing thought of they 00:55:41.460 --> 00:55:45.940 align:middle line:84% could start limit pricing and cut their prices on the Boston 00:55:45.940 --> 00:55:48.850 align:middle line:84% to Nashville route to signal to Southwest that they have 00:55:48.850 --> 00:55:51.910 align:middle line:84% low cost and tell Southwest, you should stay out 00:55:51.910 --> 00:55:55.510 align:middle line:90% of Boston Nashville. 00:55:55.510 --> 00:55:58.270 align:middle line:84% And the nice thing about it is you can look at, 00:55:58.270 --> 00:56:02.740 align:middle line:84% do they do this five months before Southwest is even 00:56:02.740 --> 00:56:04.550 align:middle line:90% operating Nashville to Miami? 00:56:04.550 --> 00:56:07.552 align:middle line:84% So you've got this long period before anything has actually 00:56:07.552 --> 00:56:09.010 align:middle line:84% changed in the airline market where 00:56:09.010 --> 00:56:11.450 align:middle line:84% it's just the threat of entry has changed. 00:56:11.450 --> 00:56:16.240 align:middle line:84% And we look at with that threat of entry changing, 00:56:16.240 --> 00:56:19.660 align:middle line:84% does United start changing its pricing, 00:56:19.660 --> 00:56:22.270 align:middle line:84% and changing price in a way that looks like it's limit pricing 00:56:22.270 --> 00:56:23.562 align:middle line:90% or something to keep entry out? 00:56:23.562 --> 00:56:26.650 align:middle line:90% 00:56:26.650 --> 00:56:28.690 align:middle line:84% Anyway, the interesting results in the paper 00:56:28.690 --> 00:56:32.570 align:middle line:84% are that this does seem to go on. 00:56:32.570 --> 00:56:35.270 align:middle line:84% So as they said, Southwest typically 00:56:35.270 --> 00:56:39.830 align:middle line:84% announces it's coming into an end point airport like five 00:56:39.830 --> 00:56:41.270 align:middle line:90% months before it does. 00:56:41.270 --> 00:56:42.800 align:middle line:90% And if you look at-- 00:56:42.800 --> 00:56:44.900 align:middle line:84% so in this case, Southwest has announced 00:56:44.900 --> 00:56:47.570 align:middle line:90% that it's coming in at time T0. 00:56:47.570 --> 00:56:51.772 align:middle line:84% And what you find is that eight months before, you 00:56:51.772 --> 00:56:52.730 align:middle line:90% don't pick up anything. 00:56:52.730 --> 00:56:56.540 align:middle line:84% But then seven, six, five months before, you pick up the airlines 00:56:56.540 --> 00:57:00.140 align:middle line:84% having lowered their prices by 7%. 00:57:00.140 --> 00:57:01.710 align:middle line:90% It goes up to 10%. 00:57:01.710 --> 00:57:03.680 align:middle line:84% And then by the time Southwest is 00:57:03.680 --> 00:57:07.250 align:middle line:84% almost in Nashville, United's prices end up 13% 00:57:07.250 --> 00:57:09.990 align:middle line:84% lower than they were before the entry occurred. 00:57:09.990 --> 00:57:13.080 align:middle line:84% And again, this is entry not competing directly on the route. 00:57:13.080 --> 00:57:15.660 align:middle line:84% This is just threat of entry causes 00:57:15.660 --> 00:57:18.470 align:middle line:90% this 13% decline in prices. 00:57:18.470 --> 00:57:22.880 align:middle line:84% And prices are then even lower after Southwest 00:57:22.880 --> 00:57:26.090 align:middle line:84% gets to Nashville, and then lower still if Southwest 00:57:26.090 --> 00:57:28.940 align:middle line:84% actually comes in and starts competing with United on Boston 00:57:28.940 --> 00:57:30.530 align:middle line:90% to Nashville. 00:57:30.530 --> 00:57:36.240 align:middle line:84% But what they're finding is that there 00:57:36.240 --> 00:57:39.030 align:middle line:84% are these price declines well before anything has actually 00:57:39.030 --> 00:57:43.290 align:middle line:84% changed in the markets other than the future entry 00:57:43.290 --> 00:57:44.700 align:middle line:90% possibility. 00:57:44.700 --> 00:57:47.520 align:middle line:84% So I think it's a striking finding. 00:57:47.520 --> 00:57:51.330 align:middle line:84% It's not perfectly clean as a reduced form prediction of they 00:57:51.330 --> 00:57:54.690 align:middle line:84% must be behaving strategically because this 00:57:54.690 --> 00:57:57.630 align:middle line:84% is a model that would have a competition effect. 00:57:57.630 --> 00:58:04.680 align:middle line:84% If United knows that Southwest is coming into the market, 00:58:04.680 --> 00:58:08.130 align:middle line:84% then it knows that it's going to be in a duopoly instead 00:58:08.130 --> 00:58:09.910 align:middle line:90% of a monopoly in the future. 00:58:09.910 --> 00:58:11.580 align:middle line:84% It, therefore, might want to behave 00:58:11.580 --> 00:58:13.500 align:middle line:84% in a way that's good for it in a duopoly 00:58:13.500 --> 00:58:15.670 align:middle line:84% relative to what's good for it in a monopoly. 00:58:15.670 --> 00:58:20.047 align:middle line:84% Now, my first thought would be, actually, that one could even 00:58:20.047 --> 00:58:20.880 align:middle line:90% go in the other way. 00:58:20.880 --> 00:58:22.980 align:middle line:84% And it could be they want to raise their prices because 00:58:22.980 --> 00:58:24.063 align:middle line:90% of the competition effect. 00:58:24.063 --> 00:58:26.070 align:middle line:84% Because if they've been trying to build customer 00:58:26.070 --> 00:58:28.620 align:middle line:84% loyalty over the years, and build frequent flyer miles, 00:58:28.620 --> 00:58:32.305 align:middle line:84% and get people to spend a lot of money, it could be that, 00:58:32.305 --> 00:58:33.930 align:middle line:84% again, you just start eat your loyalty. 00:58:33.930 --> 00:58:35.870 align:middle line:84% It's like, OK, people have good feelings about us. 00:58:35.870 --> 00:58:37.840 align:middle line:84% Good feelings are going to be worthless to us. 00:58:37.840 --> 00:58:40.007 align:middle line:84% Let's just go ahead and make all the money we want-- 00:58:40.007 --> 00:58:41.683 align:middle line:90% $2,000 to fly to Nashville. 00:58:41.683 --> 00:58:42.850 align:middle line:90% We'll take all our goodwill. 00:58:42.850 --> 00:58:44.392 align:middle line:84% We'll burn it in the next six months, 00:58:44.392 --> 00:58:45.810 align:middle line:90% and then be done with it. 00:58:45.810 --> 00:58:50.340 align:middle line:84% But it could also go the other way if, by keeping prices low, 00:58:50.340 --> 00:58:51.970 align:middle line:84% today people are loyal to United, 00:58:51.970 --> 00:58:53.850 align:middle line:84% and don't ever try Southwest, and find out 00:58:53.850 --> 00:58:56.410 align:middle line:84% what it's like to fly without having a reserved seat. 00:58:56.410 --> 00:58:58.710 align:middle line:84% And maybe you want to build up the customer loyalty 00:58:58.710 --> 00:59:00.400 align:middle line:90% and not get them to try-- 00:59:00.400 --> 00:59:03.100 align:middle line:84% get them so they don't feel like they're not angry and wanting 00:59:03.100 --> 00:59:04.438 align:middle line:90% to try Southwest. 00:59:04.438 --> 00:59:05.980 align:middle line:84% But anyway, and then a second problem 00:59:05.980 --> 00:59:10.220 align:middle line:84% is that the pre-entry stuff, I think, is really nice. 00:59:10.220 --> 00:59:13.960 align:middle line:84% The post-entry estimates, there's always this concern that 00:59:13.960 --> 00:59:19.300 align:middle line:84% while Southwest is not flying Boston to-- 00:59:19.300 --> 00:59:21.520 align:middle line:84% they eliminate markets where Southwest is flying-- 00:59:21.520 --> 00:59:24.300 align:middle line:90% 00:59:24.300 --> 00:59:27.440 align:middle line:84% where Southwest announces Nashville to Boston, 00:59:27.440 --> 00:59:30.550 align:middle line:84% they also would eliminate if Southwest announces 00:59:30.550 --> 00:59:33.520 align:middle line:84% Nashville to Miami, and then they already 00:59:33.520 --> 00:59:35.230 align:middle line:90% fly Miami to Boston. 00:59:35.230 --> 00:59:37.540 align:middle line:84% But they don't eliminate it if you 00:59:37.540 --> 00:59:41.650 align:middle line:84% can go from Nashville to Chicago, and from Chicago 00:59:41.650 --> 00:59:43.660 align:middle line:84% to Washington, and then from Washington 00:59:43.660 --> 00:59:45.340 align:middle line:90% to Boston on Southwest. 00:59:45.340 --> 00:59:48.220 align:middle line:84% Those markets, they don't eliminate. 00:59:48.220 --> 00:59:51.320 align:middle line:84% People who fly Southwest sometimes do things like that. 00:59:51.320 --> 00:59:54.620 align:middle line:84% So it could be there is actually some mismeasured real 00:59:54.620 --> 00:59:56.620 align:middle line:84% competition there that causes from people taking 00:59:56.620 --> 00:59:59.590 align:middle line:84% these multi-leg Southwest journeys to compete 00:59:59.590 --> 01:00:05.400 align:middle line:84% with the Boston Nashville flights that United has. 01:00:05.400 --> 01:00:07.140 align:middle line:90% Any questions? 01:00:07.140 --> 01:00:09.780 align:middle line:84% So the thing I wanted to finish off today 01:00:09.780 --> 01:00:14.880 align:middle line:84% is talk about, I think, a very nice recent paper by Sweeting, 01:00:14.880 --> 01:00:17.820 align:middle line:84% Roberts, and Gedge on dynamic limit pricing in the airline 01:00:17.820 --> 01:00:18.700 align:middle line:90% industry. 01:00:18.700 --> 01:00:22.935 align:middle line:84% And this is very much, again, I think I say over 01:00:22.935 --> 01:00:25.560 align:middle line:84% and over again, you don't have to be really original to succeed 01:00:25.560 --> 01:00:26.520 align:middle line:90% in economics. 01:00:26.520 --> 01:00:30.220 align:middle line:84% You can just take a good paper and do it better. 01:00:30.220 --> 01:00:32.550 align:middle line:84% And so what Sweeting, and Roberts, and Gedge are doing 01:00:32.550 --> 01:00:38.570 align:middle line:84% is looking at pricing by incumbents when Southwest 01:00:38.570 --> 01:00:41.750 align:middle line:84% is threatening entry, and trying to understand, 01:00:41.750 --> 01:00:44.990 align:middle line:84% does the limit pricing story really hang together 01:00:44.990 --> 01:00:47.270 align:middle line:84% to explain what we see in the pricing in the months 01:00:47.270 --> 01:00:48.925 align:middle line:84% before Southwest enters a market? 01:00:48.925 --> 01:00:52.573 align:middle line:90% 01:00:52.573 --> 01:00:53.990 align:middle line:84% So what they do is they start out, 01:00:53.990 --> 01:00:57.230 align:middle line:84% they discuss a theoretical model in which the limit pricing would 01:00:57.230 --> 01:00:59.090 align:middle line:90% make sense. 01:00:59.090 --> 01:01:01.580 align:middle line:84% They do a test of reduced form predictions 01:01:01.580 --> 01:01:04.760 align:middle line:84% of that model versus a non-strategic model, 01:01:04.760 --> 01:01:07.410 align:middle line:84% and then they do a structural estimation. 01:01:07.410 --> 01:01:09.470 align:middle line:84% Are the magnitudes here are sufficient to explain 01:01:09.470 --> 01:01:13.800 align:middle line:84% those 13% price drops that Goolsbee Syverson reported? 01:01:13.800 --> 01:01:16.100 align:middle line:90% So anyway, what's their model? 01:01:16.100 --> 01:01:21.320 align:middle line:84% So it's a multi-period Milgrom-Roberts limit pricing 01:01:21.320 --> 01:01:23.310 align:middle line:84% through signaling your cost model. 01:01:23.310 --> 01:01:26.570 align:middle line:84% So the incumbent has some marginal cost 01:01:26.570 --> 01:01:31.130 align:middle line:84% of carrying passengers, CIT, and it follows some Markov process 01:01:31.130 --> 01:01:34.190 align:middle line:84% on an interval, C lower bar, C upper bar. 01:01:34.190 --> 01:01:37.190 align:middle line:84% And we're going to assume it's a highly persistent Markov 01:01:37.190 --> 01:01:40.710 align:middle line:84% process, like an AR1 process with a coefficient close to 1. 01:01:40.710 --> 01:01:45.120 align:middle line:84% So costs are following an AR1 process on this interval. 01:01:45.120 --> 01:01:48.770 align:middle line:84% So at every point in time, you're always 01:01:48.770 --> 01:01:51.950 align:middle line:84% uncertain about what United's cost is because it's always 01:01:51.950 --> 01:01:54.800 align:middle line:90% different, and it's evolving. 01:01:54.800 --> 01:01:59.910 align:middle line:84% Southwest has a fixed entry cost KT, and it's drawn i.i.d. 01:01:59.910 --> 01:02:02.500 align:middle line:84% from some density in every period. 01:02:02.500 --> 01:02:05.160 align:middle line:84% And so there can be periods when it's good for Southwest 01:02:05.160 --> 01:02:07.120 align:middle line:84% to enter and bad for Southwest to enter. 01:02:07.120 --> 01:02:08.787 align:middle line:84% They're going to be more likely to enter 01:02:08.787 --> 01:02:11.280 align:middle line:90% when their costs are low. 01:02:11.280 --> 01:02:16.740 align:middle line:84% And so at each time, T, the incumbent chooses some price. 01:02:16.740 --> 01:02:19.700 align:middle line:90% It earns some profit. 01:02:19.700 --> 01:02:22.610 align:middle line:84% Southwest observes a price they charge, but doesn't 01:02:22.610 --> 01:02:24.910 align:middle line:90% observe the profit or the cost. 01:02:24.910 --> 01:02:27.440 align:middle line:84% And then Southwest learns what its cost 01:02:27.440 --> 01:02:31.230 align:middle line:84% draw is for entry costs, and then it decides to enter or not. 01:02:31.230 --> 01:02:35.120 align:middle line:84% And if it enters, there's duopoly competition 01:02:35.120 --> 01:02:39.380 align:middle line:84% with now commonly observed costs in all future periods. 01:02:39.380 --> 01:02:41.670 align:middle line:84% And firms just play the static Nash equilibrium 01:02:41.670 --> 01:02:43.200 align:middle line:90% of all future periods. 01:02:43.200 --> 01:02:45.560 align:middle line:84% If Southwest doesn't enter, then they just wait, 01:02:45.560 --> 01:02:47.060 align:middle line:84% and they can enter again next period 01:02:47.060 --> 01:02:48.540 align:middle line:90% if they get a good cost draw. 01:02:48.540 --> 01:02:51.350 align:middle line:84% So notice that Southwest, even if a market is profitable, 01:02:51.350 --> 01:02:53.510 align:middle line:84% may want to wait some number of periods 01:02:53.510 --> 01:02:56.125 align:middle line:84% to come in to wait for a good cost draw. 01:02:56.125 --> 01:02:59.473 align:middle line:90% 01:02:59.473 --> 01:03:00.890 align:middle line:84% And obviously, they're going to be 01:03:00.890 --> 01:03:04.760 align:middle line:84% more likely to come in if they think that the incumbent's cost 01:03:04.760 --> 01:03:05.340 align:middle line:90% is high. 01:03:05.340 --> 01:03:11.810 align:middle line:84% This is a model where the incumbent's cost is bouncing 01:03:11.810 --> 01:03:15.080 align:middle line:84% around on some interval, C lower bar C upper bar, 01:03:15.080 --> 01:03:18.680 align:middle line:84% and Southwest just has some other different costs, which I 01:03:18.680 --> 01:03:21.470 align:middle line:90% think we think of as down here. 01:03:21.470 --> 01:03:23.780 align:middle line:84% So we think of Southwest as being a lower cost 01:03:23.780 --> 01:03:24.860 align:middle line:90% carrier than United. 01:03:24.860 --> 01:03:27.965 align:middle line:84% So it's such that if United's cost is down here, 01:03:27.965 --> 01:03:29.840 align:middle line:84% Southwest is going to say, our cost advantage 01:03:29.840 --> 01:03:32.465 align:middle line:84% isn't big enough to pay all the fixed costs of buying the gates 01:03:32.465 --> 01:03:33.510 align:middle line:90% and hiring the employees. 01:03:33.510 --> 01:03:35.892 align:middle line:84% If United's cost is up here, Southwest 01:03:35.892 --> 01:03:37.975 align:middle line:84% is going to want to come in and compete with them. 01:03:37.975 --> 01:03:41.224 align:middle line:90% 01:03:41.224 --> 01:03:44.320 align:middle line:84% And this Markov process, like Milgrom and Roberts 01:03:44.320 --> 01:03:47.830 align:middle line:84% was this one time, you price low to signal your low cost, 01:03:47.830 --> 01:03:50.270 align:middle line:84% and then they learn your cost, and then they enter or not. 01:03:50.270 --> 01:03:52.298 align:middle line:84% Here, there's just this perpetual incentive 01:03:52.298 --> 01:03:54.340 align:middle line:84% to always signal your cost because there's always 01:03:54.340 --> 01:03:56.660 align:middle line:90% a new innovation to your cost. 01:03:56.660 --> 01:04:00.073 align:middle line:84% And that creates this incentive to continually limit 01:04:00.073 --> 01:04:01.990 align:middle line:84% price, limit price, limit price to always make 01:04:01.990 --> 01:04:03.240 align:middle line:90% them think your cost is low. 01:04:03.240 --> 01:04:11.540 align:middle line:90% 01:04:11.540 --> 01:04:14.650 align:middle line:84% So the first thing they do in the paper is-- 01:04:14.650 --> 01:04:17.980 align:middle line:84% it's a non-monotonicity test, and it's motivated 01:04:17.980 --> 01:04:20.410 align:middle line:90% by the limit pricing model. 01:04:20.410 --> 01:04:23.590 align:middle line:84% The model predicts that limit pricing 01:04:23.590 --> 01:04:27.730 align:middle line:84% is going to be most important in markets 01:04:27.730 --> 01:04:30.050 align:middle line:84% that have an intermediate entry probability. 01:04:30.050 --> 01:04:33.883 align:middle line:84% If you have a market where there is a very low entry probability, 01:04:33.883 --> 01:04:35.800 align:middle line:84% then you're not going to bother limit pricing. 01:04:35.800 --> 01:04:37.758 align:middle line:84% If there's a market where there's a very high-- 01:04:37.758 --> 01:04:44.730 align:middle line:84% so the service from Boston to Billings, Montana, 01:04:44.730 --> 01:04:47.570 align:middle line:84% there's no way Southwest is going to start flying Boston 01:04:47.570 --> 01:04:48.800 align:middle line:90% to Billings, Montana. 01:04:48.800 --> 01:04:51.980 align:middle line:84% It's just not something of the kind of market they serve. 01:04:51.980 --> 01:04:55.280 align:middle line:84% There, you don't need to limit price. 01:04:55.280 --> 01:04:57.165 align:middle line:84% If they've just entered Las Vegas 01:04:57.165 --> 01:04:59.540 align:middle line:84% and they're not serving it yet, you know sooner or later, 01:04:59.540 --> 01:05:02.323 align:middle line:84% there's going to be a Boston to Las Vegas flight on Southwest. 01:05:02.323 --> 01:05:03.990 align:middle line:84% There's nothing you can do to stop that, 01:05:03.990 --> 01:05:05.700 align:middle line:84% so you don't try to limit price there. 01:05:05.700 --> 01:05:08.630 align:middle line:84% But maybe Boston to Nashville is an intermediate sized market 01:05:08.630 --> 01:05:12.230 align:middle line:84% where you could try to make that one unattractive. 01:05:12.230 --> 01:05:15.140 align:middle line:84% And so what they're going to say is, do incumbents 01:05:15.140 --> 01:05:18.020 align:middle line:84% drop their prices most in markets 01:05:18.020 --> 01:05:21.390 align:middle line:84% with an intermediate entry probability? 01:05:21.390 --> 01:05:26.310 align:middle line:84% And so this is pretty simple, again, econometric exercise. 01:05:26.310 --> 01:05:28.070 align:middle line:84% They just have a first stage regression 01:05:28.070 --> 01:05:31.400 align:middle line:84% where they regress the probability of Southwest entry 01:05:31.400 --> 01:05:38.180 align:middle line:84% on market size, concentration of whether the endpoints have 01:05:38.180 --> 01:05:42.210 align:middle line:84% dominant carriers or not, how it fits with Southwest's network, 01:05:42.210 --> 01:05:43.968 align:middle line:90% other explanatory variables. 01:05:43.968 --> 01:05:45.510 align:middle line:84% So they have a first stage regression 01:05:45.510 --> 01:05:48.460 align:middle line:84% trying to understand where it is that Southwest enters. 01:05:48.460 --> 01:05:54.570 align:middle line:84% And then the second stage, so PM hat is the predicted entry 01:05:54.570 --> 01:05:55.920 align:middle line:90% probability. 01:05:55.920 --> 01:05:57.820 align:middle line:84% And so what they're looking for is this-- 01:05:57.820 --> 01:06:00.250 align:middle line:84% I think this sounds for Southwest post entry. 01:06:00.250 --> 01:06:06.360 align:middle line:84% They're looking at is price being reduced after entry 01:06:06.360 --> 01:06:08.782 align:middle line:84% in markets where there's an intermediate entry 01:06:08.782 --> 01:06:10.740 align:middle line:84% probability which you would get from the beta 1 01:06:10.740 --> 01:06:13.115 align:middle line:84% being negative and the beta 2 being positive? 01:06:13.115 --> 01:06:14.490 align:middle line:84% It would be saying that dips down 01:06:14.490 --> 01:06:17.610 align:middle line:84% and has low prices in those intermediate sized markets. 01:06:17.610 --> 01:06:20.643 align:middle line:84% And so they just estimate this entry probability 01:06:20.643 --> 01:06:22.560 align:middle line:84% based on market characteristics, and then they 01:06:22.560 --> 01:06:27.040 align:middle line:84% look at how the prices change related to that. 01:06:27.040 --> 01:06:34.210 align:middle line:84% And pretty clean results that those 13% price declines 01:06:34.210 --> 01:06:39.760 align:middle line:84% are relatively larger in the intermediate entry probability 01:06:39.760 --> 01:06:42.110 align:middle line:90% markets. 01:06:42.110 --> 01:06:43.490 align:middle line:90% So what you get is-- 01:06:43.490 --> 01:06:46.340 align:middle line:84% and they give many different versions of this, 01:06:46.340 --> 01:06:49.190 align:middle line:84% but the simplest one is just log of price. 01:06:49.190 --> 01:06:53.180 align:middle line:84% You regress-- This is the change in log price after entry, 01:06:53.180 --> 01:06:57.590 align:middle line:84% and you see that its interaction with the probability of entry 01:06:57.590 --> 01:07:02.300 align:middle line:84% is a negative coefficient on the probability of entry, 01:07:02.300 --> 01:07:05.610 align:middle line:84% a positive coefficient on the probability of entry squared. 01:07:05.610 --> 01:07:07.430 align:middle line:84% So it's as if that effect is something 01:07:07.430 --> 01:07:09.790 align:middle line:84% that goes down like that and then goes up like that. 01:07:09.790 --> 01:07:12.630 align:middle line:90% 01:07:12.630 --> 01:07:14.640 align:middle line:84% Another thing they do, just on the quantity side 01:07:14.640 --> 01:07:16.182 align:middle line:84% instead of on the price side, is look 01:07:16.182 --> 01:07:17.980 align:middle line:90% at how full are the airplanes. 01:07:17.980 --> 01:07:20.250 align:middle line:84% Are they sending off full or empty airplanes? 01:07:20.250 --> 01:07:22.350 align:middle line:84% And again, this one, as you would expect, 01:07:22.350 --> 01:07:24.210 align:middle line:90% it goes the opposite way. 01:07:24.210 --> 01:07:27.480 align:middle line:90% So this is the price. 01:07:27.480 --> 01:07:29.460 align:middle line:84% If you look at how full the airplanes 01:07:29.460 --> 01:07:32.370 align:middle line:84% are, that one goes up and then down so 01:07:32.370 --> 01:07:35.790 align:middle line:84% that it's your more filling up the planes 01:07:35.790 --> 01:07:37.227 align:middle line:84% in the intermediate sized markets, 01:07:37.227 --> 01:07:39.935 align:middle line:84% which presumably you fill up the planes by charging lower prices. 01:07:39.935 --> 01:07:42.490 align:middle line:90% 01:07:42.490 --> 01:07:44.590 align:middle line:84% And all the others seem to go same way. 01:07:44.590 --> 01:07:46.090 align:middle line:84% But the story seems to be that, yes, 01:07:46.090 --> 01:07:53.140 align:middle line:84% it is the intermediate sized markets where you're 01:07:53.140 --> 01:07:54.760 align:middle line:84% unsure if they're going to enter, 01:07:54.760 --> 01:07:58.820 align:middle line:84% where the price declines are biggest. 01:07:58.820 --> 01:08:00.920 align:middle line:84% And then the final thing they try to do 01:08:00.920 --> 01:08:03.960 align:middle line:90% is a structural analysis. 01:08:03.960 --> 01:08:06.780 align:middle line:84% So what they're trying to do is take that Markov model 01:08:06.780 --> 01:08:11.030 align:middle line:84% they wrote down, calibrate that model, 01:08:11.030 --> 01:08:13.615 align:middle line:84% and see whether that model explains what it has. 01:08:13.615 --> 01:08:17.120 align:middle line:90% 01:08:17.120 --> 01:08:20.090 align:middle line:84% So in particular, what they're often trying to do here 01:08:20.090 --> 01:08:26.770 align:middle line:84% is trying to not use the data in the threat period, 01:08:26.770 --> 01:08:31.630 align:middle line:84% calibrate the model using data from either the pre-period 01:08:31.630 --> 01:08:36.910 align:middle line:84% before Southwest announced, or after the Southwest entry, 01:08:36.910 --> 01:08:40.130 align:middle line:84% and then show that we can calibrate this model. 01:08:40.130 --> 01:08:42.910 align:middle line:84% And that model actually does predict price declines close 01:08:42.910 --> 01:08:45.170 align:middle line:90% to the actual price declines. 01:08:45.170 --> 01:08:49.062 align:middle line:84% So it's an exercise of saying, we're going to take this model. 01:08:49.062 --> 01:08:51.520 align:middle line:84% We're going to calibrate it using the data that we have not 01:08:51.520 --> 01:08:53.109 align:middle line:84% from that strategic investment period, 01:08:53.109 --> 01:08:54.550 align:middle line:84% and then say, what does it predict 01:08:54.550 --> 01:08:56.899 align:middle line:84% should happen in the strategic investment period? 01:08:56.899 --> 01:09:00.040 align:middle line:84% And yes, it does seem to predict the data pretty well. 01:09:00.040 --> 01:09:01.600 align:middle line:84% And then once they have it, they also 01:09:01.600 --> 01:09:04.029 align:middle line:84% use it to discuss some welfare questions 01:09:04.029 --> 01:09:07.090 align:middle line:84% because it's going to be a model that you get back consumer 01:09:07.090 --> 01:09:08.819 align:middle line:90% utility and prices out of it. 01:09:08.819 --> 01:09:11.319 align:middle line:84% You can estimate deadweight losses, and see how they change, 01:09:11.319 --> 01:09:13.694 align:middle line:84% and talk about the welfare effects of this limit pricing. 01:09:13.694 --> 01:09:15.927 align:middle line:90% 01:09:15.927 --> 01:09:17.130 align:middle line:90% So how is this done? 01:09:17.130 --> 01:09:22.500 align:middle line:84% So they restrict themselves to 109 markets. 01:09:22.500 --> 01:09:25.160 align:middle line:84% So these are 109 city pair markets, 01:09:25.160 --> 01:09:27.229 align:middle line:84% and they're all doing-- to get as close as 01:09:27.229 --> 01:09:28.854 align:middle line:84% possible to this monopoly case, they're 01:09:28.854 --> 01:09:31.040 align:middle line:84% doing 109 markets where there's one carrier that 01:09:31.040 --> 01:09:36.170 align:middle line:84% has a dominant share before Southwest comes in. 01:09:36.170 --> 01:09:38.880 align:middle line:84% Many markets, there are multiple carriers serving them. 01:09:38.880 --> 01:09:41.990 align:middle line:84% But if you pick smaller city pair markets, a lot of them 01:09:41.990 --> 01:09:44.262 align:middle line:84% that have one carrier doing most of the traffic. 01:09:44.262 --> 01:09:46.220 align:middle line:84% So you can think of it as there's a monopolist, 01:09:46.220 --> 01:09:50.170 align:middle line:84% and then the monopolist might be competing with Southwest. 01:09:50.170 --> 01:09:54.400 align:middle line:84% What they assume is that demand is a nested logit model where 01:09:54.400 --> 01:09:57.400 align:middle line:84% you have the outside good of not flying in one nest, 01:09:57.400 --> 01:10:02.260 align:middle line:84% and then Southwest versus United in the inside nest. 01:10:02.260 --> 01:10:07.120 align:middle line:84% And then they estimate that model using data 01:10:07.120 --> 01:10:14.890 align:middle line:84% from before there's any threat, and using data 01:10:14.890 --> 01:10:16.265 align:middle line:84% from after Southwest has entered. 01:10:16.265 --> 01:10:18.182 align:middle line:84% And then you're just observing the competition 01:10:18.182 --> 01:10:20.030 align:middle line:84% between the other airline and Southwest. 01:10:20.030 --> 01:10:24.280 align:middle line:84% And where do they get the IVs to do this? 01:10:24.280 --> 01:10:27.460 align:middle line:84% In the pre-period, It's fairly easy. 01:10:27.460 --> 01:10:31.120 align:middle line:84% In the pre-period, they're getting the IV for your price 01:10:31.120 --> 01:10:33.200 align:middle line:90% from fuel prices. 01:10:33.200 --> 01:10:35.320 align:middle line:84% When fuel prices go up, airlines raise prices. 01:10:35.320 --> 01:10:38.120 align:middle line:84% When fuel prices go down, airlines lower prices. 01:10:38.120 --> 01:10:41.020 align:middle line:84% Looking at how demand changes with fuel prices, 01:10:41.020 --> 01:10:43.240 align:middle line:84% they estimate substitution between the inside good 01:10:43.240 --> 01:10:44.830 align:middle line:90% and the outside good. 01:10:44.830 --> 01:10:48.310 align:middle line:84% And then for the price competition 01:10:48.310 --> 01:10:53.420 align:middle line:84% between Southwest and United, they're not as great endpoints, 01:10:53.420 --> 01:10:56.593 align:middle line:84% but it's like, using United's endpoint shares, 01:10:56.593 --> 01:10:58.760 align:middle line:84% and where United has bigger shares at the endpoints, 01:10:58.760 --> 01:11:00.390 align:middle line:84% they're going to tend to set higher prices. 01:11:00.390 --> 01:11:02.840 align:middle line:84% That's going to change the relative prices of Southwest 01:11:02.840 --> 01:11:05.390 align:middle line:90% and United. 01:11:05.390 --> 01:11:08.570 align:middle line:84% Then the marginal cost, going back 01:11:08.570 --> 01:11:15.350 align:middle line:84% to what Tobias lectures, if you first estimate the demand 01:11:15.350 --> 01:11:18.315 align:middle line:84% elasticities, and then you assume that firms 01:11:18.315 --> 01:11:19.940 align:middle line:84% are pricing optimally, you can back out 01:11:19.940 --> 01:11:21.543 align:middle line:90% what the marginal costs are. 01:11:21.543 --> 01:11:23.960 align:middle line:84% So they do that kind of approach where they first estimate 01:11:23.960 --> 01:11:25.230 align:middle line:90% the demand elasticities. 01:11:25.230 --> 01:11:27.230 align:middle line:84% Then given the demand elasticities, 01:11:27.230 --> 01:11:29.960 align:middle line:84% assume price firms are pricing at the static optimum. 01:11:29.960 --> 01:11:33.190 align:middle line:84% They back out what the costs are. 01:11:33.190 --> 01:11:37.010 align:middle line:84% And then once they get a cost period by period, 01:11:37.010 --> 01:11:38.980 align:middle line:84% they just run regressions on them 01:11:38.980 --> 01:11:40.870 align:middle line:84% to estimate AR1 coefficients and find 01:11:40.870 --> 01:11:43.270 align:middle line:90% if they're serially correlated. 01:11:43.270 --> 01:11:48.370 align:middle line:84% And then for entry costs, they just 01:11:48.370 --> 01:11:51.430 align:middle line:84% take a subset of the data using 20% of the markets, 01:11:51.430 --> 01:11:54.460 align:middle line:84% and they just say what distribution of entry costs-- 01:11:54.460 --> 01:11:57.460 align:middle line:84% given the demand estimates that we have and the profit function 01:11:57.460 --> 01:12:00.340 align:middle line:84% we've estimated for Southwest, what entry costs would 01:12:00.340 --> 01:12:02.140 align:middle line:84% rationalize the observed level of entry 01:12:02.140 --> 01:12:04.900 align:middle line:84% and the observed correlation between entry and market size? 01:12:04.900 --> 01:12:07.420 align:middle line:84% And they estimate the entry cost distributions 01:12:07.420 --> 01:12:08.950 align:middle line:90% to match that to the data. 01:12:08.950 --> 01:12:11.830 align:middle line:84% So that's where the primitives come from. 01:12:11.830 --> 01:12:13.450 align:middle line:90% What do they look like? 01:12:13.450 --> 01:12:15.700 align:middle line:84% OK, so here's what they look like. 01:12:15.700 --> 01:12:17.770 align:middle line:84% The demand estimates find that these goods 01:12:17.770 --> 01:12:19.840 align:middle line:90% are close substitutes. 01:12:19.840 --> 01:12:25.300 align:middle line:84% After entry, the average own price elasticity for incumbents 01:12:25.300 --> 01:12:27.730 align:middle line:90% is minus 2.9. 01:12:27.730 --> 01:12:30.980 align:middle line:84% So you're still going to get 30% markups over marginal cost, 01:12:30.980 --> 01:12:34.120 align:middle line:84% but demand is fairly elastic after people 01:12:34.120 --> 01:12:37.300 align:middle line:90% start competing with Southwest. 01:12:37.300 --> 01:12:39.670 align:middle line:84% The marginal cost estimates suggest 01:12:39.670 --> 01:12:43.270 align:middle line:84% that the traditional carriers are much more expensive, much 01:12:43.270 --> 01:12:46.300 align:middle line:84% higher costs of operation than Southwest does. 01:12:46.300 --> 01:12:50.140 align:middle line:84% The marginal costs are averaging 258-- 01:12:50.140 --> 01:12:53.470 align:middle line:84% the backed out marginal cost average 258 for the incumbents, 01:12:53.470 --> 01:12:56.330 align:middle line:90% and 168 for Southwest. 01:12:56.330 --> 01:12:59.274 align:middle line:84% So it looks like Southwest has a big cost advantage, which 01:12:59.274 --> 01:13:00.857 align:middle line:84% reflected comes out in the data of you 01:13:00.857 --> 01:13:03.280 align:middle line:84% see Southwest just pricing much lower 01:13:03.280 --> 01:13:06.160 align:middle line:90% than their traditional rivals. 01:13:06.160 --> 01:13:12.010 align:middle line:84% And when you do the serial correlation of the costs, 01:13:12.010 --> 01:13:14.800 align:middle line:84% it looks like costs are very highly serially 01:13:14.800 --> 01:13:15.945 align:middle line:90% correlated over time. 01:13:15.945 --> 01:13:19.490 align:middle line:90% 01:13:19.490 --> 01:13:21.440 align:middle line:84% And now on the calibrated entry costs, 01:13:21.440 --> 01:13:25.586 align:middle line:84% they just say that we can match the-- 01:13:25.586 --> 01:13:28.970 align:middle line:84% this graph here is putting two things in the paper. 01:13:28.970 --> 01:13:32.720 align:middle line:84% On the x-axis they've gotten the-- 01:13:32.720 --> 01:13:36.890 align:middle line:84% x-axis is the predicted entry probability 01:13:36.890 --> 01:13:41.180 align:middle line:84% from a regression of entry on market characteristics. 01:13:41.180 --> 01:13:43.670 align:middle line:84% The y-axis is the predicted entry cost 01:13:43.670 --> 01:13:46.730 align:middle line:84% that comes out of the theoretical model with the entry 01:13:46.730 --> 01:13:49.220 align:middle line:84% cost they've calibrated to make the two curves as close 01:13:49.220 --> 01:13:50.670 align:middle line:90% to each other as possible. 01:13:50.670 --> 01:13:53.570 align:middle line:84% And they notice that they can make these two curves look 01:13:53.570 --> 01:13:56.660 align:middle line:84% very, very close by picking an appropriate entry cost 01:13:56.660 --> 01:13:58.850 align:middle line:90% distribution. 01:13:58.850 --> 01:14:01.850 align:middle line:84% So it's both measuring the-- you have 01:14:01.850 --> 01:14:04.490 align:middle line:84% to match both the level of entry and the correlation 01:14:04.490 --> 01:14:06.710 align:middle line:84% between entry and market size, but it looks 01:14:06.710 --> 01:14:07.990 align:middle line:90% like they're able to do that. 01:14:07.990 --> 01:14:14.220 align:middle line:90% 01:14:14.220 --> 01:14:19.240 align:middle line:84% So two things they then do with the calibrated model. 01:14:19.240 --> 01:14:22.290 align:middle line:90% One is, does this model-- 01:14:22.290 --> 01:14:27.185 align:middle line:84% is the limit pricing story explaining 01:14:27.185 --> 01:14:30.120 align:middle line:84% how much limit pricing is going on in the data? 01:14:30.120 --> 01:14:33.690 align:middle line:90% So panel B here is the data. 01:14:33.690 --> 01:14:36.570 align:middle line:90% 01:14:36.570 --> 01:14:39.090 align:middle line:84% How much are incumbents reducing price 01:14:39.090 --> 01:14:42.450 align:middle line:84% as a function of the probability that entry will occur 01:14:42.450 --> 01:14:44.580 align:middle line:90% within the first four periods? 01:14:44.580 --> 01:14:47.460 align:middle line:84% And here, I guess you can see the data 01:14:47.460 --> 01:14:52.620 align:middle line:84% in theirs must apparently be not quite as large 01:14:52.620 --> 01:14:54.030 align:middle line:90% a price reduction as-- 01:14:54.030 --> 01:14:56.400 align:middle line:84% maybe it's fine if these are things. 01:14:56.400 --> 01:15:01.980 align:middle line:84% It's like, we're getting in these intermediate size markets. 01:15:01.980 --> 01:15:05.700 align:middle line:84% We're getting 13% or 14% price reductions. 01:15:05.700 --> 01:15:08.380 align:middle line:84% In these markets with very low entry probabilities, 01:15:08.380 --> 01:15:11.050 align:middle line:84% we're getting like 4% price reductions. 01:15:11.050 --> 01:15:12.840 align:middle line:90% And then in very big-- 01:15:12.840 --> 01:15:15.580 align:middle line:84% in markets where there's almost sure entry, although there 01:15:15.580 --> 01:15:19.930 align:middle line:84% are very few such markets, they're getting price increases. 01:15:19.930 --> 01:15:23.000 align:middle line:84% So that's what the data look like. 01:15:23.000 --> 01:15:26.820 align:middle line:84% And then this is what the model predicts. 01:15:26.820 --> 01:15:29.970 align:middle line:84% The model predicts that in very small markets 01:15:29.970 --> 01:15:33.600 align:middle line:84% you would have very, very tiny price cuts-- just a couple 01:15:33.600 --> 01:15:34.590 align:middle line:90% percent. 01:15:34.590 --> 01:15:36.330 align:middle line:84% In the intermediate sized markets, 01:15:36.330 --> 01:15:39.190 align:middle line:90% you would have 17% price cuts. 01:15:39.190 --> 01:15:40.860 align:middle line:84% And then as we go to this extreme, 01:15:40.860 --> 01:15:43.230 align:middle line:90% we're getting 9% price cuts. 01:15:43.230 --> 01:15:48.620 align:middle line:84% So again, it's not like the model is exactly fitting 01:15:48.620 --> 01:15:50.330 align:middle line:90% the data, but the model is-- 01:15:50.330 --> 01:15:54.080 align:middle line:84% it's the model that they've estimated not using the pricing 01:15:54.080 --> 01:15:56.522 align:middle line:84% data from the period when the threat of entry occurred. 01:15:56.522 --> 01:15:57.980 align:middle line:84% And this is what the model predicts 01:15:57.980 --> 01:15:59.848 align:middle line:84% will happen in that threat period. 01:15:59.848 --> 01:16:01.890 align:middle line:84% And this is what did happen in the threat period, 01:16:01.890 --> 01:16:04.958 align:middle line:84% and they argue that it seems like quantitatively, the model 01:16:04.958 --> 01:16:07.250 align:middle line:84% is doing a pretty good job of explaining how much limit 01:16:07.250 --> 01:16:09.390 align:middle line:90% pricing there is. 01:16:09.390 --> 01:16:13.530 align:middle line:84% And so we'll say, they think the limit pricing is explaining why 01:16:13.530 --> 01:16:15.420 align:middle line:84% this is-- the signaling is explaining-- 01:16:15.420 --> 01:16:17.362 align:middle line:90% could be explaining-- 01:16:17.362 --> 01:16:19.320 align:middle line:84% let me say, signaling is of the right magnitude 01:16:19.320 --> 01:16:23.271 align:middle line:84% to explain why this is happening. 01:16:23.271 --> 01:16:24.660 align:middle line:84% And then the second thing they do 01:16:24.660 --> 01:16:27.580 align:middle line:84% is talk about what are the welfare effects. 01:16:27.580 --> 01:16:32.905 align:middle line:84% And I guess, not surprisingly, limit pricing 01:16:32.905 --> 01:16:34.530 align:middle line:84% is really good here compared to a model 01:16:34.530 --> 01:16:35.760 align:middle line:90% with known incumbent costs. 01:16:35.760 --> 01:16:39.860 align:middle line:84% Because those limit pricing models are like this-- 01:16:39.860 --> 01:16:43.290 align:middle line:90% 01:16:43.290 --> 01:16:46.590 align:middle line:84% if you think about relative to complete information, 01:16:46.590 --> 01:16:48.660 align:middle line:84% firms are pricing a lot lower, and they're not 01:16:48.660 --> 01:16:51.010 align:middle line:90% affecting the entry much. 01:16:51.010 --> 01:17:01.050 align:middle line:84% And so you know by pricing lower, 01:17:01.050 --> 01:17:03.810 align:middle line:84% there's a substantial increase in consumer surplus 01:17:03.810 --> 01:17:05.440 align:middle line:90% from the limit pricing. 01:17:05.440 --> 01:17:08.680 align:middle line:84% There's a loss in incumbent profits from the limit pricing, 01:17:08.680 --> 01:17:11.290 align:middle line:84% but essentially, we're eliminating the deadweight loss. 01:17:11.290 --> 01:17:13.020 align:middle line:84% And so by limiting the deadweight loss, 01:17:13.020 --> 01:17:16.010 align:middle line:90% we're making things much better. 01:17:16.010 --> 01:17:18.510 align:middle line:84% And the only place where they argue that this doesn't happen 01:17:18.510 --> 01:17:22.390 align:middle line:84% is in the very large markets where there's 01:17:22.390 --> 01:17:23.980 align:middle line:90% much limit pricing going on. 01:17:23.980 --> 01:17:27.910 align:middle line:84% There, there's not really much of a welfare gain 01:17:27.910 --> 01:17:30.890 align:middle line:84% because the incumbents don't really do much limit pricing. 01:17:30.890 --> 01:17:33.560 align:middle line:84% But in general, the limit pricing just lowers prices. 01:17:33.560 --> 01:17:35.320 align:middle line:84% You have these consumer preferences where 01:17:35.320 --> 01:17:36.950 align:middle line:90% that reduces deadweight loss. 01:17:36.950 --> 01:17:39.795 align:middle line:84% This limit pricing is a socially beneficial thing. 01:17:39.795 --> 01:17:46.300 align:middle line:90% 01:17:46.300 --> 01:17:48.130 align:middle line:90% OK, questions? 01:17:48.130 --> 01:17:49.170 align:middle line:90% No? 01:17:49.170 --> 01:17:49.670 align:middle line:90% OK. 01:17:49.670 --> 01:17:55.680 align:middle line:84% So I just want to say, very nice structural paper 01:17:55.680 --> 01:17:59.720 align:middle line:90% on entry deterrence. 01:17:59.720 --> 01:18:02.160 align:middle line:84% Stephen Ryan has a very nice paper. 01:18:02.160 --> 01:18:04.920 align:middle line:84% I'm leaving it for [14.]273 because it uses a lot 01:18:04.920 --> 01:18:07.900 align:middle line:84% of methodology that's covered in [14.]273. 01:18:07.900 --> 01:18:10.470 align:middle line:84% But the basic thing of Stephen's paper 01:18:10.470 --> 01:18:18.458 align:middle line:84% is looking at when you raise environmental regulations, one 01:18:18.458 --> 01:18:20.250 align:middle line:84% of the costs the environmental regulations, 01:18:20.250 --> 01:18:21.780 align:middle line:84% is they make entry more difficult. 01:18:21.780 --> 01:18:25.440 align:middle line:84% But then the incumbents don't have to strategically deter 01:18:25.440 --> 01:18:28.620 align:middle line:84% entry anymore, so you can get much higher pricing 01:18:28.620 --> 01:18:33.330 align:middle line:84% as an additional cost of environmental regulation. 01:18:33.330 --> 01:18:36.430 align:middle line:84% Anyway, that's covered well in [14.]273. 01:18:36.430 --> 01:18:39.990 align:middle line:84% So next week, again, might shift gears. 01:18:39.990 --> 01:18:43.060 align:middle line:84% Next week, I'm going to talk about bounded rationality. 01:18:43.060 --> 01:18:45.960 align:middle line:84% Not 100% sure what I'm going do with those two lectures. 01:18:45.960 --> 01:18:47.860 align:middle line:90% I may mix them up somewhat. 01:18:47.860 --> 01:18:50.220 align:middle line:84% But anyway, next week, I'll be doing something talking 01:18:50.220 --> 01:18:52.880 align:middle line:90% about bounded rationality. 01:18:52.880 --> 01:19:06.000 align:middle line:90%